1

SMALL STREAMS BIOCRITERIA DEVELOPMENT

May 29th, 2009

NORTH CAROLINA DEPARTMENT OF ENVIRONMENT AND NATURAL RESOURCES

DIVISION OF WATER QUALITY ENVIRONMENTAL SCIENCES SECTION

THIS REPORT HAS BEEN APPROVED FOR RELEASE:

______________________________________

Jimmie Overton Chief, Environmental Sciences Section

DATE:___________________________________

2

May 29, 2009 MEMORANDUM TO: Jimmie Overton THROUGH: Trish MacPherson FROM: Eric Fleek SUBJECT: Benthic Biocriteria for the Small Streams of the North Carolina Mountains and

Piedmont: Small Streams Study (2005-2007). (1.0) Introduction Current North Carolina Division Biological Assessment Unit (BAU) Standard Operating Procedures (SOPs; NCDWQ 2006) prohibit the assignment of bioclassifications (i.e., Excellent, Good, Good-Fair, Fair, or Poor) to perennial streams with drainage areas less than 3 mi2. This deficiency stems from the fact that current bioclassification thresholds are predominately based on data obtained from streams with drainage areas greater than 3.0 mi2

(Lenat 1993). The only instance in which a stream with a drainage area equal to or less than 3.0 mi2 can receive a bioclassification pertains to Small High Quality Mountain Streams (SHQMS) which (after correction factors are applied) can receive one of the five tiered hierarchal bioclassifications (i.e., Poor, Fair, Good-Fair, Good, or Excellent; NCDWQ 2006). Other than this exception, current benthic criteria (NCDWQ 2006) limits the assignment of these bioclassifications to perennial streams with drainage areas equal to or less than 3.0 mi2 as only Not Rated or Not Impaired (NCDWQ 2006). However, in the last several years, as a result of increased sampling requests associated with the Ecosystem Enhancement Program (EEP), Total Maximum Daily Load (TMDL) studies, Outstanding Resource Waters (ORW)/High Quality Waters (HQW) reclassification studies, and NCDWQ Regional Office requests (e.g., enforcement actions), the Biological Assessment Unit (BAU) has been more frequently tasked with sampling streams with drainage areas equal to or less than 3.0 mi2. The inability to assign one of the five bioclassifications to these small streams can in certain instances decrease the usefulness of this data to end-point users. For example, a Not Impaired rating would not be (in most instances) a sufficient bioclassification for a reclassification of a waterbody to HQW or ORW. Aside from these important practical management applications associated with this resource, its has been estimated that for a given drainage network small streams (i.e., 1st, and 2nd order) can comprise up to 95% of the total stream channels present (Leopold 1956, Leopold et al. 1964) and 70-80% of total stream length (Leopold 1956, Leopold et al. 1964, Rheinhardt et al. 1999, 2005). Moreover, a recent report published by American Rivers (www.americanrivers.org/site/DocServer/Stream_Miles_Table_FINAL__2_.pdf?docID=4081) ranks North Carolina first among all 50 states in terms of total stream miles (242,691), first in terms of total perennial stream miles (123,772), and fourth in terms of total non-perennial stream miles (118,918). Additional work has shown that small first order headwater streams can

3

Figure 1. Small Streams Study Sites: Level III Ecoregions (2005-2007).

4

Figure 2. Small Stream Study Sites: Level IV Ecoregions (2005-2007).

5

contribute up to 70% of the mean-annual water volume and 65% of the nitrogen removal to second order streams and 55% of the mean-annual water volume and 40% nitrogen removal to third and higher ordered waterbodies (Alexander et al. 2007) and disproportionately affect water quality in downstream rivers and estuaries (Brinson 1993) since disturbed headwater reaches are in closest contact with sources of non-point pollution (Rheinhardt et al. 1999, 2005). Indeed, non-point source pollution is the major factor affecting poor water quality in most streams and rivers in the agricultural landscape of the southeastern U.S. Coastal Plain (Lowrance et al. 1997). In addition, small streams directly, and positively influence the biodiversity of larger river systems by acting as organic matter sources (Wipfli et al. 2007) as well as diversity sources through mechanisms such as providing unique habitat niches, providing refuge from extremes in both water temperature and discharge, and by providing refugia from competitors, predators, and exotic species (Meyer et al. 2007). The disproportionate influences that small headwater streams exert on a watershed in total magnifies the importance for the need of developing biocriteria whereby these small streams can be assigned water quality bioclassifications thus improving their management potential. A total of 122 small streams in 25 counties, spanning eight river basins, two level III ecoregions (Figure 1) and nine Level IV ecoregions (Figure 2; Griffith et al. 2002) were sampled between April, May, and June of 2005, 2006, and 2007 and initially included three coarse-scaled, land use-derived disturbance classes: “Reference” sites, “Severe Impact” sites and “Intermediate Impact” sites (sensu Whittier et al. 2007). These coarse-scaled disturbance classes were then subsequently subdivided (again, based on landuse data) into five, fine-scaled disturbance classes (i.e., “Developed”, “Mostly Developed”, “Mixed”, “Mostly Forest”, “Forest”; sensu Bryce et al. 1999, Rheinhardt et al. 2007). This known landuse, impact/test-reference/control based study design has been widely used for both deriving benthic macroinvertebrate biocriteria (Karr et al., 1986, Weisberg et al. 1997, sensu Eaton 2001, Weigel 2002) as well as for measuring the response of invertebrate communities to known disturbance gradients (Bryce et al. 1999, Cuffney et al. 2000, Tate et al. 2005, Herbst and Silldorff 2006, Kratzer et al., 2006, Smith and Lamp 2007, Carlisle et al. 2008). The analysis of this dataset was used to establish a five-tiered bioclassification threshold hierarchy for streams with drainage areas less than or equal to 3.0mi2 in both the Mountain and Piedmont level III ecoregions of North Carolina (Griffith et al. 2002). (2.0) Methods: Site Selection “Severe Impact” Site Selection. Severe impact disturbance class sites from both the Mountain and Piedmont level III ecoregions were selected a priori and were targeted towards watersheds known to be influenced by several anthropogenic stressors (Cuffney et al. 2000, sensu Eaton 2001, Carlisle et al. 2008). GIS-landuse analysis (2001 National Land Cover Data; NLCD) was conducted using ArcGIS 9.2 (Spatial Analyst—Zonal Statistics Tools) to confirm the validity and accuracy of these selections (Bryce et al. 1999, Cuffney et al. 2000, Maloney et al. 2002, Weigel 2002, Tate et al. 2005, Kratzer et al. 2006, Smith and Lamp 2007, Carlisle et al. 2008). The study sites for the severe impact dataset were selected from catchments composed primarily of a high percentage of impervious surface (i.e., urban—expressed as percent developed for this study), or from catchments with a moderate amount of impervious surface

6

(i.e., suburban), or a combination of the two. In addition, some severe impact sites from both the Piedmont and Mountains included catchments composed of some much smaller percentages of agricultural, pasture/grass, and forest land cover (Figure 3). For the Mountain severe impact sites, percent developed and percent agriculture exceeded suggested minimum thresholds (greater than 25% urban, and less than 33% agriculture) previously established for urbanized Appalachian streams (Bryce et al., 1999, Carlisle et. al. 2008) and Piedmont severe impact sites also met these criteria. “Intermediate Impact” Site Selection. Intermediate impact disturbance class sites from both the Mountains and Piedmont level III ecoregions were selected a priori and were targeted towards watersheds known to be influenced to varying degrees by anthropogenic activity (Cuffney et al. 2000, Weigel 2002, Tate et al. 2005, Kratzer et al. 2006, Smith and Lamp 2007, Carlisle et al. 2008). GIS-landuse analysis (2001 National Land Cover Data; NLCD) was conducted using ArcGIS 9.2 (Spatial Analyst—Zonal Statistics Tools) to confirm the validity and accuracy of these selections (Cuffney et al. 2000, Maloney et al. 2002, Tate et al. 2005, Carlisle et al. 2008). Intermediate impact sites from both the Mountains and Piedmont were selected from catchments comprised of a mix of agriculture and forest with slightly lesser amounts of percent developed use (Figure 3). Based on 2001 landuse data, the a priori selection process for the Mountain sites resulted in a suite of locations that, if percent agriculture and percent grass/pasture are combined, were comparable to the minimum thresholds (greater than 10% urban, greater than 33% agriculture; Figure 3) established for “mixed” impact Appalachian streams (Bryce et al. 1999, Carlisle et al. 2008), and the intermediate impact sites for the Mountains roughly fell between values for severe impact and reference Mountain sites (Figures 3). Conversely, the Piedmont study sites had a greater proportion of (combined) agricultural and grass/pasture with slightly lesser amounts of developed landuse present relative to the Mountain sites (Figure 3) but were still approximately comparable to the thresholds suggested by Carlisle et al. 2008. “Reference” Site Selection. Reference sites from both the Mountains and Piedmont level III ecoregions were selected a priori and were targeted from catchments known to be largely protected from most sources of anthropogenic activity (Cuffney et al. 2001, Tate et al. 2005). GIS-landuse analysis (2001 National Land Cover Data; NLCD) was conducted using ArcGIS 9.2 (Spatial Analyst—Zonal Statistics Tools) to confirm the validity and accuracy of these selections (Cuffney et al. 2000, Maloney et al. 2002, Tate et al. 2005, Kratzer et al. 2006, Smith and Lamp 2007, Carlisle et al. 2008). Reference sites from both the Mountains and Piedmont were selected from catchments that were almost entirely protected. These sites included watersheds contained within State Parks, National Forests, State Forests, State Game Lands, and protected lands associated with Fort Bragg. Sites that did not have catchments contained within protected natural areas were selected on the basis of a predominately forested catchment, the presence of no more than a total of two road crossings in the catchment (sensu Herbst and Silldorff, 2006, Whittier et al. 2007) and the presence of no more than two man-made farm ponds in the watershed. Based on 2001 land uses data, reference sites from the Mountains were overwhelmingly composed of forest/wetland landuse with only very minute contributions from developed, agriculture, and grass/pasture practices (Figure 3). Moreover, these sites were well under the percent developed (less than 5%) and percent agriculture (less than 50%) thresholds established by Carlisle et al.,

7

2008 for reference sites n the Appalachians and they exceeded the 80%-90% forest minima established as optimum for benthic macroinvertebrate communities (Black and Munn, 2004). Likewise, the 2001 Piedmont reference landuse data (Figure 3) were also overwhelmingly composed of forest/wetland and also met relevant criteria for reference stream selection (Bryce et al. 1999, Black and Munn 2004, Carlisle et al. 2008). However, the Piedmont reference sites had more grass/pasture and agricultural uses than did the Mountain data. The slightly higher percentages of grass/pasture and agriculture in the Piedmont were the result of less area protected by National Forests (relative to the Mountain dataset) to select sites from. Figure 3. Average Percent Landuse Comparisons Mountain (A) and Piedmont (B) Coarse-Scaled Disturbance Classes:

A. Mountain

0 20 40 60 80 100

Dis

turb

ance

Cla

ss: C

oars

e-sc

ale

B. Piedmont

Proportion (%)

0 20 40 60 80 100

Forest Water Grass/Pasture Developed Barren Agriculture

Reference

Intermediate

Severe

Reference

Intermediate

Severe

29.6

70.1

91.8

19.8

17.8

5.0

49.2

11.2

3.2

22.3

50.0

81.0

9.3

40.2

13.9

67.2

5.3

2.7

0.9

3.8

1.9

8

(3.0) Methods: Sampling Although the BAU can obtain a benthic macroinvertebrate sample at anytime of year, collections are generally concentrated during two major sampling windows. The first of these sampling windows (June-September) is typically reserved for basinwide sampling and sampling of streams greater than 3.0 mi2 in drainage area (NCDWQ 2006). The second major window is reserved for swamp sampling, which is conducted in February and March (NCDWQ, 2006). The winter sampling window is utilized for swamp streams as these waterbodies typically have no (or very little) flow during large portions of the year (NCDWQ unpublished data). In general, this low/no flow period usually occurs during the late spring and extends through summer when evapotranspiration is at its maximum. Typically, flows return to these waterbodies in mid to late fall as evapotranspiration declines. Sampling during late winter and early spring, at times of maximum baseline discharge, has been demonstrated to provide the best opportunity for detecting anthropogenic effects on invertebrate communities in these systems (NCDWQ 2006). Sampling these systems during the February-March window also offers the advantage over the summer sampling window in that confounding influences due to low dissolved oxygen, low pH, high water temperature, and decreased habitat availability due to naturally induced lower summer flows are avoided during the winter months. This is an important point as avoiding confounding effects associated with natural environmental gradients improves both the accuracy and precision of biological assessments (Carlisle et al. 2008) and is an additional argument for sampling small streams (particularly those in the level III Piedmont and level IV Slate Belt; Figure 2) during times of higher discharge (i.e., April and May) in an effort to mitigate the previously detailed confounding effects associated with swamps (i.e., lower dissolved oxygen, higher water temperature, and decreased habitat availability due to lowered summer flows) but also occasionally present in small streams. Indeed, most slate belt streams are sampled by BAU biologists during March-April for this reason. Therefore, the proposed sampling window for small streams is April, May, and June. This window is proposed for two reasons, one logistical and one for the aforementioned ecological reasons associated with confounding effects and why it is critical to avoid them (Carlisle et al. 2008). Logistically, the April-June timeframe fits between winter swamp sampling and mid-late summer (July-September) basinwide sampling. In general, personnel and property assets are substantially committed to intense field sampling during these two traditional sampling windows and accommodating additional sampling in small streams during these two major windows would prove problematic. This is particularly pronounced during the mid-late summer sampling period as that is when the majority of BAU’s benthic samples are collected. Sampling small streams in April, May, and June avoids this conflict. Future work should include repeat sampling during March from streams used in this research in an effort to extend the small streams sampling window. If March repeat sampling among a subset of sites used in this study is comparable to April, May, and June data, the sampling window could be extended (See Section 13.0). This would be logistically feasible as the vast majority of swamp sampling is traditionally completed by the end of February. In addition, re-sampling a larger subset of the streams used in this study during the month of June may also be useful in an effort to further extend the small streams sampling window towards summer. However, extending the sampling window further into the summer (e.g., July-August) would

9

likely increase the chances of naturally occurring confounding effects (particularly in the Level III Piedmont, and most notably in the level IV Slate Belt Ecoregions (NCDWQ unpublished data, Griffith et al. 2002; Figure 2). Preliminary data suggests little to no seasonal effect between April, May, and June samples. However, the June sample size is low and needs augmentation to make certain this is a valid trend for this month (See Section 11.0 for a discussion of seasonal effects). (4.0) Methods: Collection and Analysis The BAU currently maintains five sampling methods for the collection of benthic macroinvertebrates in North Carolina: Swamp Method, Full-Scale Qualitative (i.e., “Full-Scale”), EPT (Ephemeroptera, Plecoptera, Trichoptera), Boat, and the Qual-4 (NCDWQ 2006). The Swamp and Boat collection methodologies are not appropriate for small streams. This left the Qual-4, EPT, and “Full-Scale” as potential methods for sampling these systems. Given the small size and sometimes limited habitat associated with small streams, the “Full-Scale” collection method was rejected since this collection method requires three separate “bank sweeps” and two separate “riffle kicks”. During previous work in small streams, BAU staff has often found it difficult to obtain a complete sample using “Full-Scale” collection methodology due to a lack of habitat quantity. This leaves the EPT and the Qual-4 collection methodologies. In general, the EPT collection is typically reserved for streams that are greater than 3.0 mi2, or which may be habitat limited, or for sites that have no, or only low to moderate levels of anthropogenic or other physio-chemical stresses and impacts since the EPT orders are generally the least tolerant to both pollution and physical stress (Crawford and Lenat 1989, Plafkin et al. 1989, Lenat 1993, Kerans and Karr 1994, NCDWQ 2006). However, small streams by their very nature are intrinsically more susceptible to both natural and anthropogenic stresses (i.e., low summer flows, often limited habitat, less volumetric dilution to possible pollutants, etc.), and therefore the use of the EPT method in these systems may not provide enough information (i.e., EPT taxa) to make consistently accurate comparisons between sites (NCDWQ unpublished data, Lenat 1993, NCDWQ 2006). This can be a particularly intractable problem when comparing sites with severe to moderate levels of anthropogenic and/or natural stressors (NCDWQ 2006). Given these limitations, the Qual-4 method was selected for the collection of benthic macroinvertebrates for small stream bioassessment. The Qual-4 has a further advantage over the “Full-Scale” for small streams in that only one “sweep”, one “riffle kick” (plus one leaf pack and three replicates of visuals, NCDWQ 2006) is required and the Qual-4 also has an advantage over the EPT collection method in that all of the taxa collected are retained. Retaining more taxa, including many orders of facultative and pollution tolerant taxa, allows for a more complete comparison between sites with severe to moderate stressors—both natural and anthropogenic (NCDWQ unpublished data, Eaton and Lenat 1991, NCDWQ 2006). Regardless of sampling technique, there are five major benthic macroinvertebrate (BMI) community metrics that the BAU can employ in the analysis of lotic systems: Total Taxa Richness (S), EPT Taxa Richness (EPTS), EPT Abundance (EPTN), Biotic Index (BI), and EPTBI (NCDWQ 2006). EPT taxa richness (EPTS) is used with NCDWQ biocriteria to assign water quality ratings (i.e., one of the five-tiered bioclassifications, Eaton and Lenat 1991, NCDWQ 2006). EPT are generally intolerant to many types of pollution and physical stresses and higher EPT taxa richness values are indicative of better water

10

quality (Eaton and Lenat 1991). Water quality ratings can also be based on the EPTS in combination with the pollution tolerance of the macroinvertebrate community as summarized by the North Carolina Biotic Index (NCBI; Lenat 1993, NCDWQ 2006). Both tolerance values for individual species and the final biotic index values have a range of 0-10, with higher numbers indicating more tolerant species and more polluted conditions (Lenat 1993). In addition, EPTN, EPTBI, and S metrics can also be used to help examine additional between-site differences in water quality (NCDWQ 2006). In this study, each of these five metrics was evaluated to determine their relative efficacy at discriminating between streams with varying levels of disturbance. To statistically test the linear relationship between each of BAU’s five BMI metrics in response to land use type and disturbance gradient Pearson Correlation Coefficients were calculated. Since some of these correlations were statistically significant, further testing and refining of these relationships were subjected to an ANOVA calculation (F Ratio). In addition, many of the BMI metrics differ substantially in scale (e.g., BI (0.1-10.0) versus EPTN 0-270+). As a result, the Coefficient of Variation (CV) which is independent of measurement magnitude was calculated for each of the five BMI metrics to assess their respective efficacy in response to disturbance gradient. To measure the rate and magnitude of overlap among the five BMI metrics in response to disturbance gradient the Tukey-Kramer test was conducted. All statistical tests were executed in JMP 7.0 (SAS 2008). (5.0) Results and Discussion: Piedmont and Mountains (General Analysis) Benthic macroinvertebrate-based biological assessments can have multiple sources of potential variability including those arising from field sampling, laboratory sample processing, data entry, calculation of indicator variables, site assessments, as well as natural variations (e.g., seasonality and flow regime) intrinsic to biological communities (Narf et al. 1984, Diamond et al. 1996, Carter and Resh 2001, Cao et al. 2003, Clarke and Hering 2006, Clarke et al. 2006, Flotemersch et al. 2006, Herbst and Silldorf 2006, Stribling et al. 2008). Moreover, as Hynes (1970) aptly stated: “rivers are clines in the ecological sense, and such boundaries as do occur in them are ill-defined and are caused by a variety of factors.” Given this information, it is not surprising that there were some outliers (Table 1) identified from this research as well as some habitat and specific conductance variability between site groupings (Figures 4-7). In general, there are certain physical attributes that are typically required for the collection of benthic macroinvertebrates in North Carolina (NCDWQ 2006). For example, in non-swamp streams year-round flow is necessary for routine sampling and generally BAU does not sample streams with extremely low pH values (~4.0) and will not rate any stream with a pH less than or equal to 4.0 (NCDWQ 2006). During the course of this research, some samples were collected that had very low pH values (i.e., 4.0-4.8) and several of these samples had BI scores (as well as other metric scores, Table 1) that were significantly outlying from other samples within their respective disturbance classes. There were 122 total samples collected for this study. Of this total, only two of these samples (1.6%) were classified as “Invalid Outliers” (Table 1) and they were all

11

located in the Sandhills level IV ecoregions (Griffith et al. 2002). These samples were classified as invalid outliers since they were, in effect, “Invalid Samples” as they were essentially collected in violation of NCDWQ benthic macroinvertebrate pH sampling protocols (NCDWQ 2006). More importantly, they were considered invalid samples as the pH value of these two severe impact sites (Jennies Branch and UT Little River) were by far the lowest among that disturbance class as were their respective BI scores despite similar percentages of landuse present at these outlier sites compared to the remaining samples in this disturbance class (Table 1, Table 10). Neither of these invalid outlier samples were used in determining biocriteria thresholds and they were excluded from all data analysis in this report. Additional sampling efforts should be focused on obtaining data from highly impacted sites (with higher pH values) in the Sandhills level IV ecoregions (See Section 13.0).

Piedmont Excluding the two outlier sites, there were 62 streams sampled within the level III Piedmont ecoregions between April, May, and June 2005, 2006, and 2007. Ten severe impact sites were sampled with a mean drainage area of 1.9 mi2, with the largest drainage measured at 2.97 mi2 and the smallest at 0.56 mi2. Seven intermediate impact sites were sampled and mean drainage area for this disturbance class was 2.1 mi2, with the largest drainage measured at 2.7 mi2 and the smallest measured at 1.5 mi2. The remaining 45 samples were composed of reference samples and the mean drainage area for this dataset was 1.75 mi2, with the largest drainage measured at 3.0 mi2 and the smallest calculated at 0.3 mi2. Compared to the Mountain conductivity data (Figure 6) there was more overlap in stream conductivity between the three disturbance classes in the Piedmont (Figure 4) and this was predictably the most pronounced among the intermediate impact data (Figure 4). Nevertheless, the general trend was similar: conductivity was the highest among the severe impact sites and lowest among the reference sites in both ecoregions (Figure 4, Figure 6). The increased overlap of Piedmont conductivity data between disturbance classes relative to the Mountains is likely due in large part to the strong natural variations present in stream conductivity due to the effects of underlying parent geologies (NCDWQ unpublished data, Bryce et al. 1999, Jeff Reid, pers. comm., May

Table 1. Justification of “Severe Impact” Piedmont “Invalid Outliers ” Stream Disturbance

Class Justification

UT LITTLE R “Severe Impact” 1) This site had a very low pH (4.1) and was far below the average in this dataset (6.9). The 4.1 pH value was very close to the minimum pH value (4.0) that NCDWQ can use to assign bioclassifications (NCDWQ, 2006). Therefore, UT Little River should be considered an “Invalid Outlier” for this dataset and not suitable for threshold derivation. Although landuse at this site was comparable to other “Severe Impact” sites (percent developed= 57.5%, average for this dataset was 77.4%) the BI (5.13) was far lower than the average for the dataset (7.56). Additional sampling may be necessary to find severely impacted sandhills sites that have slightly higher pH values.

JENNIES BR “Severe Impact” 1) This site had a very low pH (4.4) and was far below the average in this dataset (6.9) and this value was very close to the minimum pH value (4.0) that NCDWQ can use to assign bioclassifications (NCDWQ, 2006). Therefore, Jennies Branch should be considered an “Invalid Outlier” and is not suitable for threshold derivation. Although landuse at this site was comparable to other “Severe Impact” sites (percent developed= 51.2%, average for this dataset was 57.5%) the BI (5.89) was far lower than the average for the dataset (7.56). Additional sampling may be necessary to find severely impacted sandhills sites that have slightly higher pH values.

12

2008) as well as from variation in base flow water chemistry (Feminella 2000). Furthermore, this effect is most pronounced in the Piedmont relative to the Mountain ecoregions (Jeff Reid, pers. comm., May 2008). For example, UT Bear Swamp (Piedmont reference site) had 89.91% forest/wetland land cover, 9.01% grass/pasture and 0% developed, yet the conductivity (92.2 µmhos/cm) was higher or comparable to that measured at some of the most impacted Piedmont sites (e.g., UT Mine Creek; 47 uMhos/cm, UT Richlands Creek; 94 µmhos/cm) despite the fact that both of these highly impacted sites had over 75% developed, less than 5% grass/pasture, and less than 17% forest/wetland landuse (Table 10). This illustrates the hazards of using physicochemical or landuse data alone as a method for determining effects on benthic macroinvertebrate (BMI) communities or as proxies for water quality and is illustrative of results obtained in many previous studies (Yoder and Rankin 1998, Karr and Yoder 2004, Carlisle et al., 2008). In terms of habitat scores, the Piedmont data (Figure 5) were similar to the Mountain dataset (Figures 7) in that habitat scores were somewhat more closely grouped among the three major disturbance classes than were the conductivity data between Piedmont (Figure 4) and Mountain sites (Figure 6). However, the Piedmont data (Figure 5) was less tightly grouped relative to this metric than in the Mountain ecoregions (Figure 7). Mountain There were 60 sites sampled within the level III Mountain ecoregions between April, May, and June 2005, 2006, and 2007. Ten severe impact sites were sampled and had a mean drainage area of 0.95 mi2. The largest drainage area within the severe impact disturbance class sites was 1.90 mi2 and the smallest was 0.37 mi2. Six intermediate impact sites were sampled and mean drainage area for this disturbance class was 0.97 mi2 with the largest drainage measured at 1.60 mi2 and the smallest at 0.50 mi2. The remaining 44 samples were obtained from reference sites and mean drainage area among these sites was 1.06 mi2 with the largest drainage totaling 2.70 mi2 and the smallest 0.30 mi2. With the exception of a few outlying points, conductivity was generally the highest among the severely impacted sites, lowest among the reference sites, while the conductivity among the intermediate impact study sites was generally distributed between the severe and reference streams (Figure 6). However, habitat scores (Figure 7) grouped more tightly and had slightly less overlap between disturbance classes than did the conductivity data. The slight overlap of conductivity data between disturbance classes is likely due to natural variations present in stream conductivity due to the effects of underlying parent geologies (NCDWQ unpublished data, Bryce et al. 1999, Jeff Reid, pers. comm., May 2008) as well as due to variation in base flow water chemistry (Feminella 2000). For example, Pepper Creek (reference site) had 89.7% forest/wetland land cover, 4.71% grass/pasture and only 5.58% developed (Table 11), yet the conductivity (60 µmhos/cm) was higher or comparable to that measured at some of the most impacted sites (e.g., Moore Branch; 51 µmhos/cm, UT Pigeon River; 68 µmhos/cm, King Creek 45; µmhos/cm) despite the fact that these impacted sites had over 45% developed, less than 8% grass/pasture, and less than 47% forest/wetland landuse (Table 11). Once again, this dramatically illustrates the hazards of using physicochemical or landuse parameters alone as a method for determining effects on BMI communities or as

13

water quality proxies and is consistent with previous results (Yoder and Rankin 1998, Karr and Yoder 2004, Carlisle et al. 2008).

Figure 4 and Figure 5. Piedmont Conductivity and Habitat Score by Coarse-Scale Disturbance Class.

Figure 6 and Figure 7. Mountain Conductivity and Habitat Score by Coarse-Scale

Disturbance Class.

Level-III Mountain Ecoregion Study Sites: Habitat S core by Disturbance Class

0

20

40

60

80

100

120

Individual Sites

Hab

itat S

core "Severe Impact"

"Intermediate Impact"

"Reference"

Level-III Mountain Ecoregion Study Sites: Conductiv ity by Disturbance Class

0

20

40

60

80

100

120

140

Individual Sites

Spe

cific

Con

duct

ance

(uS

/cm

)

"Severe Impact"

"Intermediate Impact"

"Reference"

Level-III Piedmont Ecoregion Study Sites: Conductiv ity by Disturbance Class

0

20

40

60

80

100

120

140

160

180

200

220

Individual Sites

Spe

ceci

fic C

ondu

ctan

ce (

uS/c

m)

"Severe Impact"

"Intermediate Impact"

"Reference"

Level-III Piedmont Ecoregion Study Sites: Habitat S core by Disturbance Class

0

20

40

60

80

100

Individual Sites

Hab

itat S

core "Severe Impact"

"Intermediate Impact"

"Reference"

14

(6.0) Results and Discussion: Piedmont and Mountain Sites (Correlation Analysis) Piedmont Overall, the mean drainage area for the 62 sites sampled in the Piedmont ecoregions was 1.81 mi2 with the largest measured at 3.0 mi2 and the smallest at 0.3mi2. As can be seen in Table 2, drainage area had no significant effect on the BI, EPTN, EPTS, or EPTBI and was only marginally correlated to S. These results are comparable to previous work in North Carolina (Lenat 1993). Among the six different landuse categories (forest/wetland, water, agriculture, barren, and grass/pasture) examined in the Piedmont, the percent developed landuse activity had overall the most powerful effect on the five benthic macroinvertebrate (BMI) metrics analyzed (Table 2). These results are consistent with prior investigations documenting the disproportionate effects of strongly deleterious landuse activities on BMI communities (Kerans and Karr 1994, Cuffney et al. 2000, Paul and Meyer 2001, Tate et al. 2005, Carlisle et al. 2008, Bressler et al. 2009). Although percent developed landuse exhibited a statistically significant influence on all five of the benthic macroinvertebrate community metrics, the BI had the strongest response and was followed by the EPTBI, EPTS, EPTN, and then the S respectively (Table 2). The percent forest/wetland landuse activity had the second most powerful overall influence on the five metrics analyzed (Table 2) and the distribution of these data are similar to prior studies that demonstrated an extremely strong negative response in BMI communities with a reduction in percent forest landuse below 70%-80% (Black and Munn 2004, Table 10, Figure 34, Figure 35). Percent forest/wetland landuse exhibited a statistically significant influence on all five of the benthic macroinvertebrate community metrics although the BI exhibited the strongest response to this landuse category and was followed by the EPTN, EPTS, and then the EPTBI, and S respectively (Table 2). Both of these results (i.e., the relative effects of percent developed and percent forest on BMI communities) are comparable with prior investigations examining the comparative impacts of strongly deleterious (e.g., developed) and positive (e.g., forested) landuse types on invertebrate communities (Kerans and Karr 1994, Cuffney et al. 2000, Paul and Meyer 2001, Carlisle et al. 2008) and the effectiveness of the BI in measuring water quality (Hilsenhoff 1987) and general stressors (Lenat 1993). Of the remaining landuse types examined in the Piedmont, only percent water had any correlation to the BI (Table 2) while percent grass/pasture, agriculture, agriculture+grass/pasture, barren, and percent water had no statistically significant effect on any of the five metrics evaluated among the Piedmont study sites (Table 2). The lack of BMI correlation to percent agriculture and grass/pasture in particular is consistent with previous findings (Meador and Goldstein 2003, Strayer et al. 2003). In addition to examining the effect of landuse on NCDWQ’s five BMI metrics, the relationship of percent silt, sand, and percent silt+sand to these metrics were also tested. For percent silt, only the EPTN demonstrated any statistically significant effect in response to percent silt in the Piedmont (Table 2). Although this correlation was marginal, it was comparable to previous results (Kondratieff and Voshell 1980, Zweig and Rabeni 2001, Thomson et al. 2004, Cover et al. 2007). All remaining BMI metrics exhibited no notable response to percent silt (Table 2). The EPTBI and EPTN, were also

15

the most sensitive to percent sand and percent silt+sand although this relationship was only moderately correlated (Table 2). Again, these results are generally similar to previous results which have documented various levels of adverse effects on invertebrate communities due to effects of fine sediment (Kondratieff and Voshell 1980, Zweig and Rabeni 2001, Thomson et al. 2004, Cover et al. 2007). The remaining BMI metrics (EPTS, BI, S) showed no statistical relationship to percent sand or percent sand+silt (Table 2). Habitat scores and specific conductance were also evaluated in this study to assess how they related to the BMI metrics as well as how these parameters correlated with landuse types. Results of this study show that habitat scores exhibited a statistically significant correlation on all five of the benthic macroinvertebrate community metrics in the Piedmont and analysis of the data demonstrate that the BI had the strongest response to habitat score and was followed by the EPTN, EPTS, EPTBI, and the S respectively (Table 2). The ability of the BI to track habitat degradation is a contrary finding from previous investigations (Kerans and Karr 1994, Shields et al. 1995, Larsen et al. 2001) but is consistent with more recent work (Bressler et al. 2009). Although all community metrics (other than S) exhibited statistically significant effects with specific conductance and is similar to recent findings (Bressler et al. 2009), the EPTBI was the most sensitive but was followed closely by the BI (Table 2). The remaining metrics were much less sensitive relative to the EPTBI and BI. In terms of the six major landuse types tested in this study, percent forest/wetland and percent developed landuse types exerted statistically significant effects on habitat scores in the Piedmont ecoregions with the strongest relationship associated with percent forest/wetland and percent developed (Table 2). These results are consistent with previous findings detailing the negative effects of developed and positive effects of forested landuse types on lotic systems (Kerans and Karr 1994, Cuffney et al. 2000, Paul and Meyer 2001, Black and Munn 2004, Carlisle et al. 2008, Bressler et al. 2009). The remaining four landuse types exhibited no statistical effect on habitat scores (Table 2). As was the case for the habitat score dataset, both percent forest/wetland and percent developed had a statistically significant effect on conductivity and the strongest relationship was found with the percent developed and percent forest/wetland respectively (Table 2) and is consistent with previous findings (Bressler et al. 2009). The remaining four land use types had no statistical effect on conductivity (Table 2). These results are consistent with previous findings detailing the biological, physical, and chemical effects of developed and forested landuse types on lotic systems (Smart et al. 1985, Kerans and Karr 1994, Herlihy et al. 1998, Bryce et al. 1999, Cuffney et al. 2000, Paul and Meyer 2001, Black and Munn 2004, Carlisle et al. 2008, Bressler et al. 2009).

16

Table 2. Piedmont Correlations: Benthic Macroinvertebrate Metrics, Landuse, Specific Conductance, and Habitat. (NOTE: BI and EPTBI scores increase with increasing impacts and are therefore positively (+) correlated to % developed. Conversely, EPTS, EPTN, and S decrease with increasing percent (%) developed landuse and are negatively (-) correlated. For percent (%) forest landuse, these correlations are reversed).

Variable by Variable Correlation Count Signif Prob Graphical Portrayal of Correlation

-1 0 +1 BI Developed % 0.8386 61 <.0001 EPTBI Developed % 0.7573 62 <.0001 EPTS Developed % -0.6808 62 <.0001 EPTN Developed % -0.6664 62 <.0001 S Developed % -0.5197 62 <.0001 BI Forest/Wetland % -0.7548 61 <.0001 EPTN Forest/Wetland % 0.7152 62 <.0001 EPTS Forest/Wetland % 0.6885 62 <.0001 EPTBI Forest/Wetland % -0.6179 62 <.0001 S Forest/Wetland % 0.4119 62 0.0009 EPTN Ag % -0.1759 62 0.1715 EPTBI Ag % -0.1491 62 0.2474 EPTS Ag % -0.1150 62 0.3733 BI Ag % -0.0664 61 0.6110 S Ag % -0.0383 62 0.7677 EPTN Grass/Pasture % -0.1878 62 0.1437 EPTS Grass/Pasture % -0.1359 62 0.2922 S Grass/Pasture % 0.0678 62 0.6007 BI Grass/Pasture % 0.0563 61 0.6662 EPTBI Grass/Pasture % -0.0180 62 0.8895 EPTN Ag+Grass/Pasture -0.2007 62 0.1178 EPTS Ag+Grass/Pasture -0.1432 62 0.2670 S Ag+Grass/Pasture 0.0553 62 0.6692 BI Ag+Grass/Pasture 0.0402 61 0.7584 EPTBI Ag+Grass/Pasture -0.0414 62 0.7492 BI Water % 0.2484 61 0.0536 EPTN Water % -0.1897 62 0.1398 EPTS Water % -0.1607 62 0.2122 EPTBI Water % 0.0816 62 0.5283 S Water % 0.0063 62 0.9611 EPTBI Barren % -0.1377 62 0.2859 BI Barren % -0.0823 61 0.5282 S Barren % -0.0477 62 0.7130 EPTS Barren % -0.0294 62 0.8209 EPTN Barren % 0.0009 62 0.9944 EPTBI % Sand -0.2908 62 0.0219 EPTN % Sand -0.2505 62 0.0495 EPTS % Sand -0.2010 62 0.1173 BI % Sand -0.0911 61 0.4850

17

Variable by Variable Correlation Count Signif Prob Graphical Portrayal of Correlation -1 0 +1

S % Sand -0.0876 62 0.4983 EPTN % Silt -0.2526 62 0.0476 EPTS % Silt -0.2063 62 0.1077 EPTBI % Silt -0.1708 62 0.1845 S % Silt -0.0824 62 0.5243 BI % Silt 0.0626 61 0.6315 EPTBI % Sand+Silt -0.2910 62 0.0218 EPTN % Sand+Silt -0.2805 62 0.0272 EPTS % Sand+Silt -0.2261 62 0.0772 S % Sand+Silt -0.0964 62 0.4559 BI % Sand+Silt -0.0586 61 0.6536 S Drainage Area 0.2550 62 0.0455 EPTBI Drainage Area 0.1753 62 0.1729 EPTN Drainage Area 0.1565 62 0.2244 EPTS Drainage Area 0.1448 62 0.2614 BI Drainage Area 0.0024 61 0.9852 Condo EPTBI 0.7756 62 <.0001 Condo BI 0.6802 61 <.0001 Condo EPTS -0.4095 62 0.0009 Condo EPTN -0.3834 62 0.0021 Condo S -0.2259 62 0.0775 Habitat BI -0.6114 61 <.0001 Habitat EPTN 0.6021 62 <.0001 Habitat EPTS 0.5478 62 <.0001 Habitat EPTBI -0.4718 62 0.0001 Habitat S 0.3506 62 0.0052 Condo Developed % 0.6546 62 <.0001 Condo Forest/Wetland % -0.4135 62 0.0008 Condo Ag % -0.2379 62 0.0626 Condo Ag+Grass/Pasture -0.2126 62 0.0971 Condo Grass/Pasture % -0.1895 62 0.1402 Condo Barren % -0.1150 62 0.3735 Condo Water % -0.0618 62 0.6330 Habitat Forest/Wetland % 0.5405 62 <.0001 Habitat Developed % -0.5144 62 <.0001 Habitat Ag % -0.1785 62 0.1650 Habitat Ag+Grass/Pasture -0.1471 62 0.2539 Habitat Grass/Pasture % -0.1285 62 0.3194 Habitat Water % 0.0631 62 0.6259 Habitat Barren % -0.0266 62 0.8371

18

Mountain The mean drainage area for the 60 samples collected in the Mountain ecoregions was 1.02 mi2 with the largest drainage area measured at 2.70mi2 and the smallest at 0.30 mi2. As can be seen in Table 3, drainage area had no significant statistical effect on the BI, EPTN, EPTS, EPTBI, or S. These findings are consistent with previous investigations conducted in North Carolina (NCDWQ unpublished data, Lenat 1993). Among the six different landuse categories (forest/wetland, developed, water, agriculture, barren, and grass/pasture) examined, the percent forest/wetland landuse activity had overall the most powerful influence on the five metrics analyzed in Mountain locations (Table 3). Moreover, the distributions are strikingly similar to prior studies which demonstrated an extremely strong negative response in invertebrate communities with a reduction in percent forest landuse below 70%-80% (Black and Munn 2004; Table 11, Figures 35, 37). Percent forest/wetland landuse exhibited a statistically significant influence on all five of the benthic macroinvertebrate community metrics although the BI had the strongest response to this landuse category followed by the EPTBI, EPTS, EPTN, and S (Table 3). Percent developed landuse exerted the second most powerful influence on the five metrics analyzed among Mountain sites (Table 3). As was the case in the Piedmont, these results are comparable with prior investigations examining the disproportionate effects of strongly deleterious landuse activities (e.g., developed) and positive (e.g., forest) landuse types on invertebrate communities (Kerans and Karr 1994, Cuffney et al. 2000, Paul and Meyer 2001, Carlisle et al. 2008, Bressler et al. 2009) and the effectiveness of the BI in measuring water quality (Hilsenhoff 1987) and general stressors (Lenat 1993). Although percent developed landuse exhibited a statistically significant influence on all five of the benthic macroinvertebrate community metrics, the BI had the strongest response followed by the EPTBI, EPTS, EPTN, and S respectively (Table 3). Percent grass/pasture landuse demonstrated the third most powerful influence on the five BMI metrics analyzed from Mountain streams. Unlike results from the Piedmont, percent grass/pasture landuse exhibited a statistically significant influence on all five of the benthic macroinvertebrate community metrics from Mountain sites (Table 3). However, evaluation of the statistical results show that the BI had the strongest response to this landuse category and was followed by the EPTBI, EPTS, EPTN, and S (Table 3). Likewise, percent agriculture in the Mountain ecoregions had a statistical effect on all five BMI metrics although it did not in the Piedmont. The statistical results indicate that the EPTBI had the strongest response to percent agriculture and was followed by the BI, EPTS, EPTN, and S (Table 3). Again, unlike the Piedmont dataset, percent agriculture + grass/pasture exhibited a statistically significant influence on all five of the benthic macroinvertebrate community metrics (Table 3). Analysis of these data demonstrates that the BI had the strongest response to this landuse combination followed by the EPTBI, EPTS, EPTN, and S (Table 3). Although percent water had no effect on any of the five BMI metrics tested, percent barren did show a borderline effect on the BI but not on any of the remaining four metrics (Table 3) and was also a result not detected in the Piedmont (Table 2). In addition to examining the effect of landuse on BAU’s five BMI metrics, the relationship of percent sand, silt, and percent silt+sand to these metrics were also tested in the Mountain ecoregions. Results of this indicate that for both percent silt, sand, and

19

sand+silt, the EPTBI exhibited the strongest relationship to these stressors, although the BI and EPTN also exhibited statistically significant effects (Table 3). These findings are similar to previous results (Kondratieff and Voshell 1980, Thomson et al. 2004, Cover et al. 2007) but are contrary to the findings of other studies which found no effect on biotic indices in response to sediment (Anagradi 1999, Zweig and Rabeni, 2001, See Section 8.0 for a more detailed discussion). The only BMI metrics that did not demonstrate any statistical significance was the S, EPTS, and EPTN for percent sand. Habitat scores and specific conductance were also evaluated in this study to assess how they related to the BMI metrics as well as how these parameters correlate with landuse types. As was the case among Piedmont streams, results in the Mountain ecoregions also demonstrated that habitat scores exhibited a statistically significant influence on all five of the BMI community metrics although the EPTBI and BI had the strongest response to habitat score (Table 3). The EPTS and EPTN were the next most sensitive to habitat score followed by the S respectively (Table 3). The ability of the BI and EPTBI to track habitat degradation is a contrary result from prior studies (Kerans and Karr 1994, Shields et al. 1995, Larsen et al. 2001) but is a similar result to more recent work (Bressler et al. 2009). Although all five metrics exhibited a statistically significant relationship to stream conductivity in the Mountains, and was a similar result from recent research (Bressler et al. 2009), the BI exhibited the strongest statistical correlation and was followed by the EPTS, EPTN, EPTBI, and S respectively (Table 3). In terms of the six major landuse types tested among Mountain streams for this study, percent forest/wetland, percent agriculture+grass/pasture, grass/pasture, developed, and agriculture all showed a statistically significant correlation to habitat score although percent forest/wetland was the most strongly correlated (Table 3). These results are consistent with previous findings detailing the biological, physical, and chemical effects of human induced landuse disruption versus non-disturbing landuse practices on lotic systems (Smart et al. 1985, Kerans and Karr 1994, Herlihy et al. 1998, Bryce et al. 1999, Cuffney et al. 2000, Paul and Meyer 2001, Black and Munn 2004, Carlisle et al. 2008, Bressler et al. 2009). Percent water did not demonstrate any statistical effect although percent barren was borderline significant (Table 3). In terms of conductivity, the Mountain data showed that the most significantly correlated variable to conductivity was percent grass/pasture, Ag+grass/pasture, forest/wetland, agriculture, and percent developed respectively (Table 3). The barren and water landuse categories exhibited no statistically significant effect on stream conductivity. Table 3. Mountain Correlations: Benthic Macroinvertebrate Metrics, Landuse, Specific Conductance, and Habitat. (NOTE: BI and EPTBI scores increase with increasing impacts and are therefore positively (+) correlated to % developed. Conversely, EPTS, EPTN, and S decrease with increasing percent (%) developed landuse and are negatively (-) correlated. For percent (%) forest landuse, these correlations are reversed).

Variable by Variable Correlation Count Signif Prob Graphical Portrayal of Correlation

-1 0 +1 BI Forest/Wetland % -0.9079 60 <.0001 EPTBI Forest/Wetland % -0.8364 60 <.0001 EPTS Forest/Wetland % 0.7983 60 <.0001

20

Variable by Variable Correlation Count Signif Prob Graphical Portrayal of Correlation -1 0 +1

EPTN Forest/Wetland % 0.7479 60 <.0001 S Forest/Wetland % 0.5941 60 <.0001 BI Developed % 0.7459 60 <.0001 EPTBI Developed % 0.6754 60 <.0001 EPTS Developed % -0.6734 60 <.0001 EPTN Developed % -0.6306 60 <.0001 S Developed % -0.4843 60 <.0001 BI Ag+Grass/Pasture 0.5964 60 <.0001 EPTBI Ag+Grass/Pasture 0.5709 60 <.0001 EPTS Ag+Grass/Pasture -0.4927 60 <.0001 EPTN Ag+Grass/Pasture -0.4627 60 0.0002 S Ag+Grass/Pasture -0.3966 60 0.0017 BI Grass/Pasture % 0.5929 60 <.0001 EPTBI Grass/Pasture % 0.5642 60 <.0001 EPTS Grass/Pasture % -0.4840 60 <.0001 EPTN Grass/Pasture % -0.4556 60 0.0003 S Grass/Pasture % -0.3897 60 0.0021 EPTBI Ag % 0.5362 60 <.0001 BI Ag % 0.5096 60 <.0001 EPTS Ag % -0.5043 60 <.0001 EPTN Ag % -0.4576 60 0.0002 S Ag % -0.4043 60 0.0014 BI Barren % 0.2349 60 0.0708 EPTN Barren % -0.2193 60 0.0923 EPTS Barren % -0.2004 60 0.1247 EPTBI Barren % 0.1929 60 0.1398 S Barren % -0.1170 60 0.3732 EPTBI Water % 0.0415 60 0.7530 BI Water % -0.0148 60 0.9108 EPTS Water % 0.0121 60 0.9268 S Water % 0.0108 60 0.9348 EPTN Water % -0.0105 60 0.9366 EPTBI %Sand+Silt 0.6416 60 <.0001 BI %Sand+Silt 0.5429 60 <.0001 EPTN %Sand+Silt -0.3366 60 0.0085 EPTS %Sand+Silt -0.3310 60 0.0098 S %Sand+Silt -0.0881 60 0.5034 EPTBI %Sand 0.5548 60 <.0001 BI %Sand 0.4693 60 0.0002 S %Sand 0.0377 60 0.7748 EPTS %Sand -0.2121 60 0.1037 EPTN %Sand -0.2086 60 0.1097 EPTBI % Silt 0.4882 60 <.0001

21

Variable by Variable Correlation Count Signif Prob Graphical Portrayal of Correlation -1 0 +1

EPTN % Silt -0.4220 60 0.0008 BI % Silt 0.4132 60 0.0010 EPTS % Silt -0.4008 60 0.0015 S % Silt -0.2962 60 0.0216 S Drainage Area 0.1835 60 0.1604 EPTS Drainage Area 0.1825 60 0.1628 EPTN Drainage Area 0.1644 60 0.2094 EPTBI Drainage Area 0.1446 60 0.2704 BI Drainage Area -0.0074 60 0.9554 BI Condo 0.6139 60 <.0001 EPTS Condo -0.5538 60 <.0001 EPTN Condo -0.5336 60 <.0001 EPTBI Condo 0.4794 60 0.0001 S Condo -0.4069 60 0.0013 EPTBI Habitat -0.7937 60 <.0001 BI Habitat -0.7683 60 <.0001 EPTS Habitat 0.7032 60 <.0001 EPTN Habitat 0.6652 60 <.0001 S Habitat 0.4900 60 <.0001 Habitat Forest/Wetland % 0.7531 60 <.0001 Habitat Ag+Grass/Pasture -0.6098 60 <.0001 Habitat Grass/Pasture % -0.6067 60 <.0001 Habitat Developed % -0.5536 60 <.0001 Habitat Ag % -0.5149 60 <.0001 Habitat Barren % -0.2358 60 0.0697 Habitat Water % 0.0347 60 0.7922 Condo Grass/Pasture % 0.6878 60 <.0001 Condo Ag+Grass/Pasture 0.6791 60 <.0001 Condo Forest/Wetland % -0.6433 60 <.0001 Condo Ag % 0.3990 60 0.0016 Condo Developed % 0.3815 60 0.0026 Condo Barren % 0.1808 60 0.1668 Condo Water % -0.0681 60 0.6054 (7.0) Results and Discussion: Piedmont and Mountain (Analysis by Disturbance Class) In terms of biocriteria selection for the severe impact disturbance class, the BI had the lowest Coefficient of Variation (CV) of all the other metrics for both the Piedmont and the Mountains (Figure 8, Figure 9). Within the intermediate impact disturbance class dataset, the BI again had a substantially lower CV relative to all the other BMI metrics in both the Piedmont and Mountains (Figure 8, Figure 9). Similarly, the reference dataset also demonstrated that the BI had the lowest CV of all the other metrics in the Piedmont and Mountains (Figure 8, Figure 9).The evaluation of this is crucial as the CV has

22

repeatedly been shown to be a highly effective tool for selecting metrics and bioassessment methods that are the most effective in discriminating between reference and impaired streams (Gebler 2004, Black et al. 2004, Herbst and Silldorff 2006, Stribling et al. 2008) and has been used frequently to compare the relative precision among various indicators, data sets, and metrics (Gomez and Gomez 1976, Diamond et al. 1996, Feminella 2000, Zweig and Rabeni, 2001, Schloesser and Nalepa 2002, Tullos et al. 2006, Rehn et al. 2007, Cuffney et al. 2007). In addition, this method is particularly useful when comparing metrics with differing units of magnitude (SAS 2006) and is an extremely useful test for comparing the raw discriminatory efficacy of a given metric (Dr. Catherine Truxillo, pers. comm., April 2008). Based on this analysis, only the BI had a CV that was small enough to provide useful information for establishing water quality biocriteria thresholds in the Piedmont and Mountains and was the least variable of all the other metrics (Figures 8-10). The only instance where the BI failed to outperform the other BMI metrics occurred when the reference foothills data was combined with the reference Mountains data (Figure 10). However, even in this solitary instance the difference in the CV between the top performing (S) metric relative to the BI (1.72 for the coarse-scale disturbance data, and 2.34 for fine-scale) were the smallest differences measured between the top performing BI and the next best CV metric for the entire Mountain dataset. The only such difference that was smaller (1.58) occurred between the BI and the EPTS from the mostly forest disturbance class (Figure 12). Indeed, the remaining differences in the CV of the BI to the next best metric ranged from a low of 2.61 to a high of 24.53 and thus solidify the overall stability of the BI relative to the EPTBI, EPTS, EPTN, and S. In addition to the CV measurement, the S, EPTS, EPTN and EPTBI demonstrated significant overlap (see Figure 12A) between all disturbance classes in both ecoregions with the largest differences measured among the coarse-scale disturbance classes (e.g., overlap between reference sites with severe impact sites as well as overlap between intermediate impact sites with severe and reference sites (Figures 13-17, Figures 23-27). In addition, the S, EPTN, EPTBI, and EPTS also exhibited significant rates of overlap between more narrowly separated, finer-scaled disturbance classes (e.g., overlap between forest and mixed, forest with mostly forest, mostly forest with mixed, as well as developed with mixed, developed with mostly developed and mostly developed with mixed; Figures 18-22, Figures 28-32) in the Piedmont and Mountains. As a result, the EPTBI, S, EPTS, and EPTN are not optimal for use in the establishment of biocriteria in streams less than or equal to 3.0 mi2 in North Carolina’s Piedmont and Mountains as these metrics consistently failed to detect both large and small differences in disturbance gradients within the benthic macroinvertebrate community. These findings are consistent with previous investigations (Herbst and Silldorff 2006, Stribling et al. 2008). These data suggest that the BI is the most effective metric for reliably and consistently discriminating between reference, intermediate impact and severe impact sites in both ecoregions (Figures 13-17, Figures 23-27), as well as discriminating between finer-scaled disturbance gradients (Figures 18-22, Figures 28-32). Indeed, only the BI had the “power” to reliably and consistently discern between reference, intermediate impact and severe impact sites in both ecoregions. Specifically, “power” is measured empirically as

23

the rate of overlap between “Test” sites (i.e., classified as severe impact, intermediate impact, developed, mostly developed, and mixed sites in this study; sensu Herbst and Silldorff, 2006) with “Control” sites (i.e., sites classified as reference, forest, and mostly forest in this study; sensu Herbst and Silldorff, 2006) while “overlap” is measured as the proportion of test sites that exceed the reference distribution values for that metric (Herbst and Silldorff, 2006). In effect, overlap evaluates the signal-to noise ratio by considering the rate and magnitude of separation between the test and reference site distributions (Herbst and Silldorff, 2006). Therefore, the lower the power and higher the rate of overlap, the more frequently one will misidentify “Test” sites as “Reference” sites and “Reference” sites with “Test” sites (Herbst and Silldorff 2006). In other words, this situation could result in misidentifying impaired streams with those in reference, or near-reference condition. In summary, the effectiveness of the BI, and relative ineffectiveness of the EPTS, EPTBI, EPTN, and S in discriminating between both large and small scale disturbance gradients in small Piedmont and Mountain streams can be seen in Figures 13-32. Combined, these data suggest that the Biotic Index (BI) is the most powerful and precise metric (among the five BMI metrics analyzed) for the establishment of biocriteria for streams with drainage areas less than or equal to 3.0 mi2 in North Carolina’s Piedmont and Mountains. In summary, the BI had no overlap between the three major (landuse derived; Figure 3, Table 10, Table 11) coarse-scale disturbance classes (i.e., severe impact, intermediate impact, reference) while the remaining four remaining metrics (EPTS, EPTN, EPTBI, and S) did demonstrate overlap (Figures 13-17, Figures 23-27, Table 10, Table 11). Importantly, when these three coarse-scale disturbance classes were further subdivided into five fine-scaled disturbance classes (also based on landuse; Figure 33, Table 10, Table 11) in the Piedmont the BI again demonstrated no overlap between forest, mostly forest, mixed, mostly developed, and developed landuse disturbance gradients (Figure 17, Figure 22, Table 10) and the BI in the Mountains also exhibited no instances of overlap between either the coarse-scale or fine-scale disturbance classes (Figure 27, Figure 32, Table 11). Indeed, despite the fact that the slightly more variable foothills reference data were included with the Mountain reference data, the BI still exhibited no overlap between the two closest reference categories (forest and mostly forest; Figure 27, Figure 32, Table 11). These results further bolster the robustness of the BI relative to the S, EPTS, EPTN, and EPTBI. In total, the BI exhibited no overlap among all of the disturbance gradients (both coarse and fine-scale) in the Piedmont and Mountain ecoregions (Figure 17, Figure 22, Figure 27, Figure 32). Conversely, the S, EPTS, EPTN, and EPTBI did exhibit overlap between not only the fine-scale disturbance class categories (Figures 18-21, Figures 28-31), but the S, EPTS, EPTN, and EPTBI demonstrated overlap between the worst, and near-worst sites (developed/mostly developed and severe impact) with the best, and near-best sites (forest/mostly forest and reference, Figures 13-16, Figures 23-26). Therefore, the EPTN, EPTBI, EPTS, and S are not optimal for use in biocriteria development in streams with drainage areas less than or equal to 3.0mi2 in the Piedmont and Mountains of North Carolina as frequent overlap of metrics between impaired (e.g., severe impact) and non-impaired (e.g., intermediate impact and reference) sites leads to misclassification of

24

reference sites with impaired sites and vice versa (Herbst and Silldorff, 2006, Stribling et al. 2008). This condition would have severe negative consequences for water resource managers (EPA 2000). Moreover, establishing a gradient of disturbance also aids in interpreting results of sampling as well as guiding the calibration of metrics composing the qualitative and/or quantitative indices of biological assemblage response (Bryce et al. 1999). Further, the comparison of these metrics to disturbance gradients tests whether a given metric’s response covers the entire spectrum of water quality, habitat quality, and which metrics are most (or least) sensitive to extremes in impact, and which taxa disappear or become more prevalent with the accumulation of stressors (Bryce et al. 1999). Therefore, testing benthic macroinvertebrate metrics against subtle disturbance gradients are crucial in selecting the metric(s) which provides the best overall signal (Bryce et al. 1999). These data demonstrate the power of the BI as a reliable method to not only discriminate between the best and worst sites but also to distinguish between varying levels of subtler land-use based disturbance gradients in both the Piedmont and Mountains. Distinguishing gradations in biological structure and function is a key underpinning of the regulatory process of assigning streams to different categories of aquatic life use attainment (Jackson 2004) and the clarity and accuracy with which different metrics and/or methods permit identification of thresholds and intermediate subdivisions of impairment is another key feature that should be considered when comparing metric and method performance (Herbst and Silldorff 2006). This is important since effective water quality management requires tools that can detect degrees and gradients of degradation to biological communities since a decision of simply “degraded” or “not degraded” can often be accurately inferred from examining discharger locations, land use trends, and previous investigations (EPA 2000, Eaton 2001). In both the Piedmont and Mountains, the S, EPTN, EPTBI, and EPTS were not as effective in detecting gradations in disturbance gradients. Conversely, the BI was highly effective at detecting gradients in disturbance. The BI also demonstrated the lowest Coefficient of Variation (CV) relative to the other four metrics in both ecoregions, which further bolsters the use of this metric in biocriteria (Gomez and Gomez 1976, Diamond et al. 1996, Herbst and Silldorff 2006, Tullos et al. 2006, SAS 2006 Dr. Catherine Truxillo, pers. comm., April 2008, Stribling et al. 2008,). In fact, the CV of the BI among the Piedmont severe impact, intermediate impact, and reference disturbance classes (Figure 8) were well within the range (10-15) of previously recommended (precision) measurement quality CV objectives (Stribling et al., 2008). All of the remaining metrics (EPTN, EPTS, EPTBI, and S) within the three major coarse-scale disturbance classes fell beyond this recommended range (Stribling et al. 2008) and without exception were all higher than the BI for each disturbance class in the Piedmont (Figure 8). In addition, when the CVs by metric are examined among the five, finer-scaled landuse categories in the Piedmont (Figure 11) the BI was lower than all the other metrics for each of the five fine-scale disturbance categories and all of the BI CV values met the recommended precision CV objective (Stribling et al. 2008). Conversely, only the EPTBI CV for the mostly developed and forest landuse categories met these thresholds (Figure 11).

25

In the Mountains, the BI CV among the severe impact, intermediate impact, and reference sites without “Foothills” streams (Figure 9; See Section 9.0 for a detailed discussion of the foothills dataset) were well within the range (10-15) of previously recommended (precision) measurement quality CV objectives (Stribling et al. 2008). Conversely, only the reference EPTBI CV fell within this range in the Mountains (Figure 9). Remaining metrics among all of the three major, coarse-scale disturbance classes fell beyond this recommended range (Stribling et al. 2008) and without exception they were all higher than the BI for each disturbance class among Mountain sites (Figure 9). Moreover, in the Mountains, the BI CV (excluding reference foothills sites) was lower than any of the other metrics within the five, finer-scaled (forest, mostly forest, mixed, mostly developed, and developed) landuse categories (Figure 12) and the BI CV (excluding the forest+foothills forest category) were well below the recommended precision (10-15) CV requirements (Stribling et al. 2008). Conversely, only the mostly developed EPTBI CV and mostly forest EPTS CV met these thresholds (Figure 12). When the foothills sites are added into the Level III Mountain ecoregions reference disturbance class dataset, in both the coarse-scale (Figure 10) and fine scale land-use derived disturbance gradients (Figure 12; See Section 9.0 for a detailed discussion of the foothills dataset), the CVs of the five BMI metrics shifted somewhat as the S (16.17) and EPTS (17.89) outperformed the BI (18.51) in the coarse-scale dataset (Figure 10), and all three CV values associated with these metrics failed to meet the minimum precision requirements (Stribling et al. 2008). However, when examined singularly, the CV values for metrics contained within just the foothills dataset show that only the BI (15.85, Figure 10) met the Stribling et al. 2008 precision thresholds (10-15) while the CVs for the S, EPTS, EPTN, and EPTBI exceeded these minimum precision values (Figure 10). While all of the foothills sites contained core Mountain taxa, some of these sites had slightly higher BI values than some of the Mountain reference streams even though the landuse trends between these two groups were comparable (Table 11). The slightly higher BI scores associated with the foothills dataset, are responsible for the increased BI CV for the combined coarse-scale and fine-scale foothills/Mountains reference dataset (Figure 10, Figure 12). Furthermore, if the foothills reference data are grouped with the Piedmont reference dataset, the BI CV of that combination increases from 12.13 to 17.77 and helps illustrate the transitional nature of the foothills data. See section 9.0 for a detailed treatment of the foothills dataset. Of all the metrics examined in the Piedmont and Mountains, the BI was the most strongly correlated to the percent forest/wetland and percent developed landuse categories (Table 2, Table 3). This is particularly significant as these two landuse types have consistently been demonstrated to exert the strongest negative (developed) and positive (forest) effect on benthic macroinvertebrate communities (Kerans and Karr 1994, Cuffney et al. 2000, Paul and Meyer 2001, Weigel 2002, Black and Munn 2004, Moore and Palmer 2005, Kratzer et al. 2006, Smith and Lamp 2007, Carlisle et al. 2008). Therefore it is crucial to select a metric that best tracks these relationships (Karr et al. 1986, Lenat 1988, Weisburg et al. 1997). Furthermore, the BI has the additional advantage over the other benthic macroinvertebrate metrics evaluated in this study as the BI has been shown to be an

26

effective measure of water quality (Hilsenhoff 1987) and of general stressors (Lenat 1993). Indeed, these two landuse types combined to have the strongest effect on all five of the benthic metrics in the Piedmont (Table 2) and Mountains (Table 3) and the percent forest/wetland and percent developed landuse types were also shown to have the strongest correspondence to habitat score and conductivity in the Piedmont (Table 2). In the Mountains, percent forest/wetland had the strongest overall correlation to habitat score although every other landuse type (excluding barren and water) were also statistically significant (Table 3). In the Mountains, the conductivity was significantly correlated to percent grass/pasture and forest followed by percent agriculture and developed (Table 3). Although some of the other non-BI metrics were also significantly related to percent developed and percent forest/wetland, these non-BI metrics, in every instance had a much higher CV than did the BI in the Piedmont (Table 2), and in all but one instance, the BI has a lower CV than all other BMI metrics in the Mountains (Table 3). In instances where parameters have both a high degree of correlation and a high CV (i.e., are inconsistent) the strength of the correlation is weakened and therefore parameters that are both highly correlated and highly consistent (i.e., have a low CV) should be given the greater weighting (Cuffney et al. 2007). This would indicate the BI is superior to the other non-BI metrics in this study. In the Mountains, percent sand and percent sand+silt had the strongest statistical effect on the EPTBI and BI respectively (Table 3), while percent silt exhibited the strongest statistical correlation to the EPTBI, EPTN, and BI (Table 3). These findings are contrary to previous work which showed that biotic indices have little or no correlation to sedimentation (Anagradi 1999, Zweig and Rabeni, 2001). However, the Anagradi study artificially manipulated sedimentation rates and therefore those results may not have been representative of natural conditions (Zweig and Rabeni, 2001). Moreover, the Zweig and Rabeni study assessed naturally sandy streams in central Missouri. These streams and their invertebrate communities are adapted to robust, naturally occurring sediment regimes and therefore would be expected to be less sensitive to small and moderate increases in sedimentation (Zweig and Rabeni, 2001). Indeed, the differential response in BMI communities to sediment from mountain areas (i.e., invertebrates sensitive to sediment) to lowlands (i.e., invertebrates less sensitive to sediment) has been demonstrated previously (Connolly and Pearson, 2007). Therefore the present findings in the Mountain ecoregion are consistent with many other studies which have demonstrated the deleterious effects of fine sediment on lotic benthic macroinvertebrates (Kondratieff and Voshell, 1980, Thomson et al. 2004, Cover et al. 2007). The fact that the BI and EPTBI were significantly correlated to sand, silt, and sand+silt in the Mountains is likely the result of the fact that the invertebrate communities in Mountain streams are not as well adapted to the higher rates of naturally occurring sedimentation in the Piedmont and are therefore more sensitive to these stressors and again is consistent with Connolly and Pearson (2007). In the Piedmont, only the EPTBI and EPTN had any significant statistical correlation to percent sand+silt and it was only marginally so (Table 2). In addition, only the EPTBI

27

and EPTN had any correlation to percent sand, while only the EPTN demonstrated any correlation to percent silt (Table 2). All remaining BMI metrics demonstrated no significant relationship to percent silt, percent sand, or percent sand+silt in the Piedmont (Table 2). Unlike the findings in the Mountains, the Piedmont data are more consistent with Anagradi 1999 and Zweig and Rabeni, 2001 who demonstrated that measures of abundance are significantly correlated to increasing rates of sedimentation among naturally sediment laden central Missouri streams. Again, the findings in the Piedmont are likely related to the higher amounts of naturally occurring sand and silt found in the level III Piedmont ecoregions and most of its level IV ecoregions relative to the level III Mountains and its associated level IV ecoregions (Griffith et al. 2002). Indeed, differences among the mean values of percent silt, sand, and sand+silt (p=0.0071, p=0.0515, and p=0.0174, respectively) between the Piedmont and Mountains were all statistically significant even though the Piedmont sites with the highest percentages of naturally occurring sand and silt (level IV Sandhills ecoregions, Griffith et al. 2002) were removed from the analysis. Had sandhills sites been included in this analysis the significance of this statistical effect would have been much larger. Sandhills Level IV ecoregions data were removed from consideration as these sites are naturally composed nearly entirely of all sand (Griffith et al. 2002) and would therefore artificially skew the analysis. In the Mountains, although the EPTBI was slightly more related to habitat score, percent sand, silt, sand+silt, and percent agriculture relative to the BI (Table 3) the EPTBI (in every instance and dataset excluding the Mountains+foothills sites) had a much higher CV than did the BI (Figure 9, Figure 10, Figure 12). As noted previously, in instances where parameters have both a high degree of correlation and a high CV (i.e., are inconsistent) the strength of the correlation is weakened and therefore parameters that are both highly correlated and highly consistent (i.e., have a low CV) should be given the greater weighting (Cuffney et al. 2007). This would indicate the BI is superior to the EPTBI in this dataset. Moreover, the difference between the EPTBI and BI among these metrics was trivial and therefore the performance of the EPTBI and BI for these variables is relatively comparable in the Mountains. For example, in the Mountain ecoregions, ANOVA (F-Ratio) scores for the EPTBI and BI were 66.07 and 61.57 for habitat, 13.71 and 12.38 for percent agriculture, 16.34 and 16.24 for percent grass/pasture, and 16.56 and 16.34 for percent grass/pasture+agriculture respectively. Conversely, of all the metrics and specifically relative to the EPTBI, the BI was the most strongly associated with percent forest/wetland and percent developed. For example, for percent forest, the ANOVA (F-Ratio) values for the BI and EPTBI respectively were 306.33 and 162.58 and for percent developed they were 85.36 and 62.50. The fact that the BI had a stronger statistical correlation to the forest/wetland and developed landuse types in both ecoregions is significant as these two landuse types have consistently been demonstrated to exert the strongest negative and/or positive effect on benthic macroinvertebrate communities (Kerans and Karr 1994, Cuffney et al. 2000, Paul and Meyer 2001, Weigel 2002, Black and Munn 2004, Moore and Palmer 2005, Kratzer et al. 2006, Smith and Lamp 2007, Carlisle et al. 2008). Therefore it is crucial to select a metric that best tracks this relationship (Karr et al. 1986, Lenat 1988, Weisburg et al. 1997). Indeed, as previously seen, these two landuse types combined to have the strongest statistical effect

28

on all five of the benthic metrics in the Piedmont and Mountains and the percent forest/wetland and percent developed landuse types were also shown to have the strongest correlation to habitat score and conductivity in the Piedmont. In the Mountains, habitat score was most strongly correlated to percent forest/wetland although every other landuse type (other than percent water) was also highly correlated (Table 3). In terms of conductivity, percent grass/pasture, percent grass/pasture+Ag, and percent forest/wetland were most correlated to conductivity although every remaining landuse type (other than barren and water) were also statistically significant (Table 3). In total, these data strongly suggest the BI is the best performing, efficacious, and precise metric for use in the establishment of biocriteria for small streams in North Carolina’s Piedmont and Mountain ecoregions. In addition, using the BI is advantageous relative to the EPTBI, EPTS, and EPTN since in extremely anthropogenically disturbed (or naturally variable) streams there will likely be a very reduced EPT fauna present since these taxa are generally the most sensitive to pollution and other forms of physicochemical stress (Crawford and Lenat 1989, Plafkin et al. 1989, Lenat 1993, Kerans and Karr 1994, NCDWQ 2006). As a result, in highly disturbed streams (either disturbed naturally or anthropogenically) there can often be zero, or at best too few EPT (and non-EPT taxa present; Maloney et al. 2002) to generate a meaningful EPTBI, EPTS, or EPTN metric and therefore comparing amongst highly (or intermediately impacted sites) using these metrics will be highly limited or effectively impossible (Crunkilton and Duchrow 1991, NCDWQ 2006). Moreover, whenever the EPTN is at or below 30 (as was the case in eight of the Piedmont sites; Table 10, and five of the Mountain sites; Table 11), the EPTBI and EPTS metrics have been found to have little interpretive meaning (NCDWQ 2006). In fact, this is the primary reason why in streams with known or suspected (natural or anthropogenic) impacts, the EPT method is generally not recommended (NCDWQ 2006). This situation is analogous to BAU’s current prohibition on sampling streams with pH values less than or equal to 4.0 as these sites will have such low EPT and total taxa richness values (as well as low EPT and non-EPT abundances) that the calculation of community metrics will largely lose meaning and context (NCDWQ 2006). This will reduce the efficacy at which these sites can be compared to not only other low pH systems, but will also limit these streams from being effectively compared to sites that have higher pH values since it may not be possible to separate the deleterious effects of low pH with those of anthropogenic impact (NCDWQ 2006). Moreover, the EPTBI and EPTS are less effective if the diversity and abundance of more tolerant invertebrate groups need to be evaluated (NCDWQ 2006). In addition, the BI has been shown to be more reliable than the EPTS metric when assessing small streams (Lenat 1993, Lenat and Barbour 1994, Stewart and Loar 1994) and biotic indices, multi-assemblage (Carlisle et al. 2008) and composite (Watzin and McIntosh 1999) measures have repeatedly demonstrated their superior utility and accuracy relative to other benthic macroinvertebrate community metrics (Chutter 1972, Jones et al. 1981, Hilsenhoff 1982, 1987, 1988, Narf et al. 1984). Moreover, biotic indices have proven to be accurate measures of water quality (Hilsenhoff 1987), ecosystem health (Karr 1991, 1993), and of general stressors (Leant 1993).

29

For example, immediately after clear-cutting, increases in EPT taxa occurred in streams draining the clear-cut catchment, compared with that of a nearby reference stream draining a protected forested catchment (Swift 1983, Wallace et al. 1996). Several EPT taxa normally confined to larger downstream segments colonized the stream draining the clear-cut watershed without a significant loss of headwater EPT taxa (Swift 1983, Wallace et al. 1996). With forest regrowth, the number of EPT taxa exhibited successive declines compared to that of the reference stream. In contrast, compared to the reference stream, the BI correctly indicated decreased biological integrity immediately following clear-cutting, which subsequently improved following forest regrowth (Swift 1983, Wallace et al. 1996). Furthermore, other types of disturbance (e.g., low intensity agriculture, riparian thinning, or pasture operations) may also inflate EPT taxa richness metrics as well as total taxa richness (S) in small streams by increasing productivity (NCDWQ unpublished data, Quinn 2000) and this response is consistent with the subsidy-stress response (Odum et al. 1979). Conversely, in these situations the BI will in most cases correctly reflect the decrease in biological integrity (NCDWQ unpublished data). Indeed, recent work in North Carolina from small headwater streams ranging in size from 14 to 816 acres (640 acres equals 1.0 mi2) has shown higher taxa richness scores in impacted catchments relative to reference watersheds due to a disproportionate increase in tolerant taxa (Burton 2004) while other studies have shown that metrics limited to only measures of EPT (e.g., EPTS, EPTN, EPTBI) are often lower in headwater streams relative to larger downstream reaches and therefore may compromise its use for biological monitoring in these smaller streams (Lenat and Barbour 1994, Stewart and Loar 1994). These data demonstrate the relative effectiveness of the BI versus measures of taxa richness for accurately assessing the integrity of benthic macroinvertebrate communities (Burton 2004) and lends additional support for its use in the establishment of biocriteria for streams with drainage areas less than or equal to 3.0 mi2 in North Carolina’s Piedmont and Mountain ecoregions.

Figure 8. Coefficient of Variation (CV): Piedmont Coarse-Scale Disturbance

Classes.

Coefficient of Variation (CV): Piedmont Coarse-Scal e Disturbance Class

0

10

20

30

40

50

60

70

80

90

Reference Intermediate Severe Impact

Coe

ffic

ient

of V

aria

tion

(CV

)

BI

EPTBI

EPTN

EPTS

S

30

Figure 9. Coefficient of Variation (CV): Mountain Coarse-Scale Disturbance

Classes.

Figure 10. Coefficient of Variation (CV): Foothills and Mountain Coarse-Scale Disturbance Classes.

Coefficient of Variation (CV): Mountain Coarse-Scal e Disturbance Class

0

10

20

30

40

50

60

70

80

Reference Int ermediate Severe Impact

Coe

ffici

ent o

f Var

iatio

n (C

V)

BI

EPTBI

EPTN

EPTS

S

Coefficient of Variation (CV): Mountain and Foothil ls Coarse-Scale Disturbance Class

0

5

10

15

20

25

30

35

Mountains Reference (No Foothills) Mountain s Reference (With Foothills) Foothills Ref erence

Co

effic

ient

of

Var

iatio

n (C

V)

BI

EPTBI

EPTN

EPTS

S

31

Figure 11. Coefficient of Variation (CV): Piedmont Fine-Scale Disturbance Classes.

Coefficient of Variation (CV): Piedmont Fine-Scale Dist urbance Class

0

10

20

30

40

50

60

70

80

90

Developed Mostly Developed M ixed Mostly Forest Forest

Coe

ffici

ent o

f Var

iatio

n (C

V)

BI

EPTBI

EPTN

EPTS

S

Figure 12. Coefficient of Variation (CV): Piedmont Fine-Scale Disturbance Classes.

1 Only one sample obtained from “Developed” disturbance class and a CV could not be calculated. Additional data from the “Developed” disturbance class will be collected spring 2009 (See Section 13.0).

Coefficient of Variation (CV): Mountains and Foothi lls Fine-Scale Disturbance Class

0

10

20

30

40

50

60

70

Developed 1 Mostly Developed Mixed Mostly Forest Forest (No Foothills) Forest (With Foothills)

Coe

ffici

ent o

f Var

iatio

n (C

V)

BI

EPTBI

EPTN

EPTS

S

32

Figure 12(A). Explanatory Graphic for Interpreting Overlap Diagrams: (Figures 13-32, Pages 33-36).

0

50

100

150

200

250

300

EP

TN

Reference Intermediate Severe All Pairs

Tukey-Kramer0.05

GroupMean

Comparison Circles

One result

angle > 90 degrees angle = 90 degrees angle < 90 degrees

not significantly

different

borderline significantly

different

significantly different

No overlap

significantly different

large samplesize = smallerdiameter circle

orgin of circle = group mean

Angle of intersection

not significantly

different

33

Figure 13. S by Coarse-Scale Disturbance Class Figure 14. EPTS by Coarse-Scale Distribution Class (PIEDMONT) (PIEDMONT)

Figure 15. EPTN by Coarse-Scale Disturbance Class Figure 16. EPTBI by Coarse-Scale Disturbance Class (PIEDMONT) (PIEDMONT)

Figure 17. BI by Coarse-Scale Disturbance Class (PIEDMONT)

0

10

20

30

40

50

No.

EP

TS

peci

es (

EP

TS

)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer0.05

0

50

100

150

200

250

300

EP

T A

bund

ance

(EP

TN

)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer0.05

0

2

4

6

8

10

EP

T B

iotic

Inde

x (E

PT

BI)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer0.05

0

2

4

6

8

10

Bio

tic In

dex

(BI)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer

0.05

0

20

40

60

80

100

No.

Spe

cies

(S

)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer0.05

34

Figure 18. S by Fine-Scale Disturbance Class Figure 19. EPTS by Fine-Scale Disturbance Class (PIEDMONT) (PIEDMONT)

Figure 20. EPTN by Fine-Scale Disturbance Class Figure 21. EPTBI by Fine-Scale Disturbance Class (PIEDMONT) (PIEDMONT)

Figure 22. BI by Fine-Scale Disturbance Class (PIEDMOMT)

20

40

60

80

100

No.

Spe

cies

(S

)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All Pairs

Tukey-Kramer

0.05

0

10

20

30

40

50

No.

EP

TS

peci

es (

EP

TS

)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All Pairs

Tukey-Kramer

0.05

0

50

100

150

200

250

300

EP

T A

bund

ance

(EP

TN

)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All Pairs

Tukey-Kramer

0.05

0

2

4

6

8

10

EP

T B

iotic

Inde

x (E

PT

BI)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All Pairs

Tukey-Kramer

0.05

0

2

4

6

8

10

Bio

tic In

dex

(BI)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All PairsTukey-Kramer

0.05

35

Figure 23. S by Coarse-Scale Disturbance Class Figure 24. EPTS by Coarse-Scale Disturbance Class (MOUNTAINS) (MOUNTAINS)

Figure 25. EPTN by Coarse Scale-Disturbance Class Figure 26. EPTBI by Coarse-Scale Disturbance Class (MOUNTAINS) (MOUNTAINS)

Figure 27. BI by Coarse-Scale Disturbance Class (MOUNTAINS)

0

20

40

60

80

100

No.

Spe

cies

(S

)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer0.05

0

10

20

30

40

50

No.

EP

TS

peci

es (

EP

TS

)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer0.05

0

50

100

150

200

250

300

EP

T A

bund

ance

(EP

TN

)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer

0.05

0

2

4

6

8

10

EP

T B

iotic

Inde

x (E

PT

BI)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer

0.05

0

2

4

6

8

10

Bio

tic In

dex

(BI)

Reference Intermediate Severe

Coarse

All Pairs

Tukey-Kramer

0.05

36

Figure 28. S by Fine-Scale Disturbance Class Figure 29. EPTS by Fine-Scale Disturbance Class

(MOUNTAINS) (MOUNTAINS)

Figure 30. EPTN by Fine-Scale Disturbance Class Figure 31. EPTBI by Fine-Scale Disturbance Class (MOUNTAINS) (MOUNTAINS)

Figure 32. BI by Fine-Scale Disturbance Class (MOUNTAINS)

0

20

40

60

80

100

No.

Spe

cies

(S

)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All Pairs

Tukey-Kramer

0.05

0

10

20

30

40

50

No.

EP

TS

peci

es (

EP

TS

)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All Pairs

Tukey-Kramer

0.05

0

50

100

150

200

250

300

EP

T A

bund

ance

(EP

TN

)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All Pairs

Tukey-Kramer0.05

0

2

4

6

8

10

EP

T B

iotic

Inde

x (E

PT

BI)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All Pairs

Tukey-Kramer

0.05

0

2

4

6

8

10

Bio

tic In

dex

(BI)

Forest Mostly Forest Mixed Mostly Developed Developed

Fine

All Pairs

Tukey-Kramer

0.05

37

(8.0) Development of Piedmont and Mountain Biocriteria Thresholds Based on this information, the BI was selected for use in the development of biocriteria and a five-tiered bioclassification hierarchy (based on the BI) and derived from the landuse data was developed from the originally (a priori) selected (coarse-scale) reference, intermediate, and severe impact study sites from both the Piedmont and Mountains (sensu Karr et al. 1986, Lenat 1988, Weisburg et al. 1997, sensu Eaton 2001, Weigel 2002, Tate et al. 2005, Kratzer et al. 2006, Smith and Lamp 2007, Carlisle et al. 2008). The sites comprising these coarse-scale disturbance classes were further subdivided into five fine-scaled disturbance categories (based on landuse data; sensu Carlisle et al. 2008, Table 10 and Table 11) and were then sorted based on percent forest, percent developed, percent grass/pasture, and percent agriculture. Percent water and percent barren were excluded from this process as these landuse types exhibited no effect on any of the five BMI metrics tested in this study. As a result of this sorting into disturbance gradients based on the landuse data, the corresponding BI scores (Table 10, Table 11) were then used as the basis for the five-tiered bioclassification thresholds (i.e., “Excellent”, “Good”, “Good-Fair”, “Fair”, and “Poor”) that NCDWQ currently uses to assign water quality (bioclassification) ratings to streams greater than 3.0 mi2 (NCDWQ 2006). These small stream thresholds are presented below for both the Piedmont (Table 4) and Mountains (Table 5). Table 4. Piedmont Biocriteria Thresholds.

Table 5. Mountain Biocriteria Thresholds.

PIEDMONT Bioclassification

BI Values Number of Samples (n)

Number of Samples Falling Outside of

Thresholds EExxcceell lleenntt </= 4.36 17 1

GGoooodd 4.37-5.48 29 0 GGoooodd--FFaaii rr 5.49-6.02 6 0

FFaaii rr 6.03-6.98 5 0 PPoooorr >/= 6.99 5 0

MOUNTAIN Bioclassification

BI Values Number of Samples (n)

Number of Samples Falling Outside of

Thresholds EExxcceell lleenntt </= 3.75 44 0

GGoooodd 3.76-4.99 6 0 GGoooodd--FFaaii rr 5.00-5.80 5 0

FFaaii rr 5.81-6.70 4 0 PPoooorr >/= 6.71 1 0

38

The results of this process indicate that for the Piedmont and the Mountains the BI is not only a powerful predictor of primarily deleterious or positive landuse types (e.g., developed and forest) with corresponding disturbance gradients (e.g., severe impact with reference, Table 10 and Table 11) but the BI is also very effective at predicting smaller gradients of response in the invertebrate assemblage that are associated with subtler (and evenly or nearly evenly mixed landuse types) and corresponding subtle disturbance gradients (Figures 13-32, Table 10 and Table 11). These findings are comparable to previous work (Kerans and Karr 1994, Cuffney et al. 2000, Paul and Meyer 2001, Black and Munn 2004, Moore and Palmer 2005, Carlisle et al. 2008) Indeed, not only were (most) of the mean landuse percentages, habitat scores, conductivity, BI (and to a lesser extent) most of the other four benthic macroinvertebrate metrics significantly different between landuse-derived disturbance gradients and associated bioclassifications (separated by two bioclassification steps) in the Piedmont (Table 6) and Mountains (Table 8) but many of these parameters also exhibited significant differences among the five (landuse-derived) fine-scale disturbance categories and associated five-tiered bioclassifications separated by only one step (Table 7, Table 9). Specifically, 72.7% of the means among the variables in the Piedmont and Mountains (separated by two bioclassification steps and disturbance classes) were significantly different (Table 6 and Table 8) while 43.2% of the means were significantly different in the Piedmont and 60.6% of the means significantly different in the Mountains (when separated by only one bioclassification and disturbance class step; Table 7, Table 9). The fact that the differences in the means of these variables were often statistically significantly between the various disturbance categories demonstrates the presence of “real” physical, chemical, and biological differences among these sites and this technique has been used previously to demonstrate the same point (Carlisle et al. 2008). In short, the statistically significant differences in the means between disturbance gradients strongly support the creation of a five-tiered bioclassification system. Conversely, if most of these parameters lacked statistically significant differences between disturbance gradients and associated bioclassifications, then a five-tiered bioclassification hierarchy would not be supported. Of particular note, relative to the S, EPTS, EPTN, and EPTBI, the BI consistently exhibited not only the strongest statistical difference (often with p values <0.0001) between the disturbance classes, but the BI was statistically significant in every instance whereas the other metrics frequently were not (Tables 6-9). . This further supports the selection of the BI as the preferred metric for use in biocriteria development in small streams.

39

Table 6. Piedmont Mean Landuse Differences by Fine-Scale Disturbance Class: Separation by Two Bioclassification Steps and Two Disturbance Classes.

Table 7. Piedmont Mean Landuse Differences by Fine-Scale Disturbance Class: Separation by One Bioclassification Step and One Disturbance Class.

Disturbance Class &

BIOCLASSIFICATION

S

EPTS

EPTN

EPTBI

BI

Habitat

Conductivity

%

Forest

%

Developed

%

Grass/Pasture

% Ag

Reference: Forest (EXCELLENT)

TO Intermediate: Mixed

Forest (GOOD-FAIR)

MEAN DIFFERENCE p=0.7309 p=0.0744 p=0.0989 p=0.0059 p<0.0001 p=0.2144 p=0.2889 p=0.0544 p=0.4557 p=0.0210 p=0.8973

Reference: Mostly Forest (GOOD)

TO Impact: Mostly

Developed (FAIR)

MEAN DIFFERENCE p=0.1817 p=0.0124 p=0.0238 p=0.0462 p<0.0050 p=0.1854 p=0.0320 p=0.0004 p=0.0008 p=0.0006 p=0.8950

Intermediate: Mixed (GOOD-FAIR)

TO Impact: Developed

(POOR)

MEAN DIFFERENCE p=0.0491 p=0.0138 p=0.0270 p=0.0381 p<0.0001 p=0.0730 p=0.0566 p=0.0251 p<0.0001 p=0.8440 p=0.0223 Significant p values highlighted in yellow. Borderline significant p values highlighted in orange.

Disturbance Class &

BIOCLASSIFICATION

S

EPTS

EPTN

EPTBI

BI

Habitat

Conductivity

%

Forest

%

Developed

%

Grass/Pasture

% Ag

Reference: Forest (EXCELLENT)

TO Reference: Mostly Forest

(GOOD)

MEAN DIFFERENCE p=0.2227 p=0.0007 p=0.0004 p=0.0032 p<0.0001 p=0.0377 p=0.0435 p=0.2162 p=0.3621 p=0.1057 p=0.4908

Reference: Mostly Forest (GOOD)

TO Intermediate: Mixed

(GOOD-FAIR)

MEAN DIFFERENCE p=0.9752 p=0.4347 p=0.2913 p=0.2992 p=0.0179 p=0.7809 p=0.2047 p=0.0092 p=0.0133 p=0.0186 p=0.0470

Intermediate: Mixed (GOOD-FAIR)

TO Impact: Mostly Developed

(FAIR)

MEAN DIFFERENCE p=0.3386 p=0.1960 p=0.1480 p=0.1042 p=0.0091 p=0.0874 p=0.0056 p=0.0856 p=0.0010 p=0.0069 p=0.1123

Impact: Mostly Developed (FAIR)

TO Impact: Developed

(POOR)

MEAN DIFFERENCE p=0.2445 p=0.2103 p=0.4021 p=0.3401 p=0.0200 p=0.5370 p=0.6945 p=0.1681 p=0.0318 p=0.2261 p=0.9502 Significant p values highlighted in yellow. Borderline significant p values highlighted in orange.

40

Table 8. Mountain Mean Landuse Differences by Fine-Scale Disturbance Class: Separation by Two Bioclassification Steps and Two Disturbance Classes.

Table 9. Mountain Mean Landuse Differences by Fine-Scale Disturbance Class: Separation by One Bioclassification Step and One Disturbance Class.

Disturbance Class &

BIOCLASSIFICATION

S

EPTS

EPTN

EPTBI

BI

Habitat

Conductance

%

Forest

%

Developed

%

Grass/Pasture

% Ag

Reference: Forest (EXCELLENT)

TO Intermediate: Mixed

Forest (GOOD-FAIR)

MEAN DIFFERENCE p=0.0928 p=0.0014 p=0.0005 p=0.0138 p<0.0001 p=0.0101 p=0.0094 p=0.0020 p=0.0139 p=0.0267 p=0.1956

Reference: Mostly Forest (GOOD)

TO Impact: Mostly

Developed (FAIR)

MEAN DIFFERENCE p=0.7314 p=0.0056 p=0.0022 p=0.0433 p=0.0070 p=0.3446 p=0.2566 p=0.0057 p=0.0813 p=0.5651 p=0.6250

Intermediate: Mixed: Developed

(GOOD-FAIR) TO

Impact: Developed (POOR)

MEAN DIFFERENCE NA1 NA NA NA NA NA NA NA NA NA NA Significant p values highlighted in yellow. Borderline significant p values highlighted in orange. 1 Not enough data to generate a meaningful t-test. See Section 13 For Further Information.

DisturbanceClass &

BIOCLASSIFICATION

S

EPTS

EPTN

EPTBI

BI

Habitat

Conductivity

%

Forest

%

Developed

%

Grass/Pasture

% Ag

Reference: Forest (EXCELLENT)

TO Reference: Mostly Forest

(GOOD)

MEAN DIFFERENCE p=0.0201 p=0.0013 p=0.0005 p=0.0009 p=0.0001 p=0.0220 p=0.1218 p=0.0042 p=0.2375 p=0.0307 p=0.0061

Reference: Mostly Forest (GOOD)

TO Intermediate: Mixed

(GOOD-FAIR)

MEAN DIFFERENCE p=0.7984 p=0.0427 p=0.0117 p=0.2599 p=0.0014 p=0.5290 p=0.1631 p=0.0105 p=0.0459 p=0.0973 p=0.5930

Intermediate: Mixed (GOOD-FAIR)

TO Impact: Mostly

Developed (FAIR)

MEAN DIFFERENCE p=0.0917 p=0.0819 p=0.0368 p=0.4716 p=0.0626 p=0.8747 p=0.1518 p=0.0155 p=0.1401 p=0.5888 p=0.9001

Impact: Mostly Developed (FAIR)

TO Impact: Developed

(POOR)

MEAN DIFFERENCE NA1 NA NA NA NA NA NA NA NA NA NA Significant p values highlighted in yellow. Borderline significant p values highlighted in orange. 1 Not enough data to generate a meaningful t-test. See Section 13 For Further Information.

41

Although there is some variation in the landuse percentages between sites grouped into a particular disturbance classes (Figure 33), the general BI trends are tightly coupled to the relative proportions of percent developed, percent forest/wetland, percent agriculture, and percent grass/pasture in both ecoregions (Figure 35, Figure 37, Table 10, Table 11). For example, based on the BI scores and resultant biocriteria thresholds, as you progress “downward” from landuse percentages associated with an Excellent bioclassification to landuse patterns corresponding to a Good bioclassification the relative proportions of percent development, agriculture, and grass/pasture will increase while relative proportions of percent forest/wetland will decrease. Interestingly, as you progress downward from a Good bioclassification to a Good-Fair bioclassification, while the percent forest/wetland continues a decreasing trend, and percent developed continues an increased trend, the proportion of their relative changes is less and this holds true for both ecoregions (Figure 35, Figure 37, Table 10 and Table 11). As the rate of change in these two landuse types lessens, the rate of change in the percent agriculture and percent grass/pasture proportionately increases. It is only once you continue “downward” from the Good-Fair bioclassification and associated landuse trends towards the Fair and Poor bioclassifications do the percent agriculture, and percent grass/pasture landuse percentages resume their downward progression, while percent forest continues to decrease and percent developed continues to increase (Figure 35, Figure 37, Table 10 and Table 11). These landuse trends are consistent with the character of the resulting land use disturbances and are reflected by the BI-derived bioclassifications. For example, BI scores associated with Excellent bioclassifications should have mostly all forest, with little to no input from other landuse types that are related to disturbance (Black and Munn 2004). Biotic index scores related to Good bioclassifications should have slightly less forest with some of the difference being increasingly comprised of landuse types associated with disturbance (e.g., developed, agriculture, and grass/pasture). BI scores related to the Good-Fair bioclassification should include quantities of landuse that have less percent developed and less percent forest categories, but rather there should be more “intermediate” effects (e.g., grass/pasture and agriculture) present (Carlisle et al. 2008). Conversely, as the BI scores increase and associated bioclassifications decrease to Fair and Poor, the landuse types that lack deleterious effects on invertebrate communities (percent forest/wetland; Black and Munn 2004), or that have lessened deleterious effects (grass/pasture and agriculture; Carlisle et al. 2008) should continue to decrease while the landuse type most deleterious to invertebrate communities (percent developed; Kerans and Karr 1994, Cuffney et al. 2000, Paul and Meyer 2001, Black and Munn 2004, Moore and Palmer 2005, Carlisle et al. 2008) should increase. As can be seen in Figure 35, Figure 37, Table 10, and Table 11, this is the trend in both ecoregions. Although the S, EPTS, EPTBI, and EPTN also tracked these trends, (Figure 34, Figure 36), their much higher CV values (Figures 8-12) and higher rates of overlap (Figure 13-14, Figure 16, Figures 23-26, Figures 28-31) between all disturbance gradients (both coarse-scale and fine-scale) make them less optimum relative to the BI for use in biocriteria development for small streams.

42

Figure 33. Average Percent Landuse Comparisons Mountain (A) and Piedmont (B)

Proportion (%)

0 20 40 60 80 100

0 20 40 60 80 100

Dis

turb

ance

Cla

ss: F

ine-

Sca

le

ForestWaterGrass/PastureDevelopedBarrenAgriculture

A. Mountain

B. Piedmont

91.9

70.1

42.7

20.7

4.9

17.8

23.2

20.4

99.0

57.6

32.6

11.2

3.1

86.6

77.7

50.0

27.7

17.0

8.3

17.3

40.2

14.2

4.3 77.5

57.0

5.3

Forest

Mostly Forest

Mixed

Mostly Developed

Developed

Forest

Mostly Forest

Mixed

Mostly Developed

Developed

3.8

2.8

2.4

1.8

2.3

43

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Reference/Forest

Reference/Mostly Forest

Intermediate Impact/Mixed

Severe Impact/MostlyDeveloped

Severe Impact/Developed

Figure 34. Piedmont Total Taxa Richness (S), EPTS, EPT Abundnace (EPTN), Habitat, Conductivity and Landuse by Fine-Scale Dis turbance Class.

S

EPTS

EPTN

Habitat

Conductivity

% Forest

% Developed

% Ag

% Grass/Pasture

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Reference/Forest

Reference/Mostly Forest

Intermediate Impact/Mixed

Severe Impact/MostlyDeveloped

Severe Impact/Developed

Figure 35. Piedmont Biotic Index (BI), EPTBI, and P ercent Landuse by Fine-Scale Disturbance Class.

BI

EPTBI

% Forest

%Developed

%Ag

%Grass/Pasture

MEANS:

MEANS:

44

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Reference/Forest

Reference/Mostly Forest

Intermediate Impact/Mixed

Severe Impact/MostlyDeveloped

Severe Impact/Developed

Figure 36. Mountain Total Taxa Richness (S), EPTS, EPT Abundance (EPTN), Habitat, Conductivity and Landuse by Fine-Scale Disturbance Class.

S

EPTS

EPTN

Habitat

Conductivity

% Forest

% Developed

% Ag

% Grass/Pasture

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Reference/Forest

Reference/Mostly Forest

Intermediate Impact/Mixed

Severe Impact/MostlyDeveloped

Severe Impact/Developed

Figure 37. Mountain Biotic Index (BI), EPTBI, and P ercent Landuse by Fine-Scale Disturbance Class.

BI

EPTBI

% Forest

%Developed

%Ag

%Grass/Pasture

MEANS:

MEANS:

45

*The Talbots Branch BI (5.57) Exceeded the Minimum Excellent Threshold for the Level III Piedmont Ecoregions (Table 4).

Table 10. Piedmont Study Sites: Raw Data. Five-Tiered Biotic Index Ranges Based on Sorting by Landuse: % Forest, % Developed, % Grass/Pasture, and % Agriculture. Note Overlap of the S, EPTS, EPTN, and EPTBI Between Disturbance Classes. Severe Impact/Developed Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % BI S EPTS EPTBI EPTN DA Condo Habitat Ut Mine Cr 16.49 1.89 4.95 75.63 0 1.04 7.12 23 2 4.7 4 0.56 47 79 Tanyard Br 26.93 0 8.57 61.79 0 2.71 7.78 15 3 6.88 14 1.16 202 46 Ut Richlands Cr 17.22 0 3.2 79.59 0 0 7.94 14 2 6.81 13 1.38 94 61 Nutbush Cr 19.91 0 4.36 74.83 0.11 0.79 7.19 49 6 7.09 28 2.28 198 58 Beaverdam Cr 4.25 0 0.22 95.53 0 0 7.79 31 3 7 16 2.97 105 71 MEANS 17.0 0.37 4.3 77.5 0.02 0.90 7.564 26 3.2 6.49 15 1.67 129.2 63 Severe Impact/Mostly Developed Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % BI S EPTS EPTBI EPTN DA Condo Habitat Ut Richlands Cr 18.06 0 42.45 38.65 0 0.84 6.32 38 6 5.98 21 0.61 138 75 Marlowe Cr 39.04 0 9.55 50.48 0.23 0.69 6.95 26 5 5.62 9 1.74 141 39 UT Muddy Fk 9.65 0 10.69 76.54 0 3.12 6.81 63 11 5.92 44 2.56 176 49 Hasketts Cr 38.51 0 5.27 56.04 0 0.17 6.98 29 1 7 3 2.9 118 61 Penwood Br 33.36 0.11 3.25 63.28 0 0 6.03 51 14 4.76 51 2.9 149 79 MEANS 27.7 0.02 14.2 57.0 0.04 0.96 6.618 41 7.4 5.85 25.6 2.14 144.4 60.6 Intermediate Impact/Mixed Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % BI S EPTS EPTBI EPTN DA Condo Habitat Ut Haw R 35.2 0.8 54.5 3.4 0.2 5.8 5.54 75 18 4.6 96 0.5 25 79 Birch Fk 33.8 0.8 56.3 2.6 0 6.4 5.76 65 20 5.4 68 2.4 70 83 Wells Cr 54.6 0.3 38.6 4.3 0.1 2 6.02 54 13 5.7 59 2.7 97 72 Brushy Cr 49.1 1.2 41.6 6.3 0 1.6 5.55 82 26 3.9 107 2.1 81 76 L Rocky Run 61.67 0 28.23 8.36 0 1.7 5.71 19 5 4.05 16 0.3 24 80 UT Nicks Cr 65.35 0.34 22.25 6.6 0 5.5 5.74 21 4 3.41 22 1 18 63 MEANS 50.0 0.57 40.2 5.3 0.05 3.8 5.72 53 14.3 4.51 61.3 1.5 52.5 75.5 Reference/Mostly Forest Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % BI S EPTS EPTBI EPTN DA Condo Habitat Ut Tar R 89.87 0 7.58 1.98 0.57 0 5.29 36 14 2.97 67 0.4 44.3 73 Talbots Br 91.49 0 6.93 0.72 0 0.87 5.57* 51 20 4.88 127 0.5 92 85 L Stonehouse Cr 73.71 0.06 20.48 3.18 0.03 2.54 5.01 61 22 3.86 89 1.2 48 59 Ut Powells Cr 93.4 0 4.79 1.01 0.02 0.77 5.08 44 14 2.85 56 1.75 113 76 Rattlesnake Cr 75.53 0.14 21.9 1.99 0 0.44 5.42 57 16 4.5 74 1.8 89 79 Parkwood Br 80.85 0.1 15.5 1.49 1.14 0.92 5.39 70 24 4.61 119 2.9 71 75 Little Rocky Cr 88.49 0 7.59 3.55 0.03 0.34 5.18 85 31 4.13 143 2.9 56.6 75 Jordan Cr 87.13 0 8.08 4.24 0.14 0.41 5.18 61 15 4.83 79 2.8 39 79 Jordan Cr 87.13 0 8.08 4.24 0.14 0.41 4.85 57 22 3.75 86 2.8 49 81 Ut L Buffalo Cr 88.53 0 9.85 1.62 0 0 5.03 52 18 4.81 82 2.9 80 86 Ut Bear Swp 89.91 0 9.01 0 0.09 0.99 4.65 58 24 4.01 130 0.7 92.2 87 N Pr Anderson Cr 11.68 0 73.31 8.56 0 6.45 4.68 54 20 3.13 75 0.88 17 62 N Pr Anderson Cr 11.68 0 73.31 8.56 0 6.5 4.42 44 12 2.92 65 0.88 15 79 Tanyard Cr 77.76 0.04 19.57 2.27 0 0.37 4.64 67 27 3.1 124 0.95 58 60 Arnett Br 90.84 0 7.1 1.9 0 0.17 4.5 64 33 3.6 141 1.2 52 86 Big Br 82.91 0 8.61 8.07 0 0.41 4.75 53 20 3.35 90 1.3 15 67 W Pr Juniper Cr 82.19 0 10.71 5.93 0 1.17 4.39 34 11 3.38 48 1.4 11 83 W Br Maclean Cr 95.28 0 4.12 0.36 0 0.24 4.45 58 23 3.57 97 1.7 47 88 W Br Maclean Cr 95.28 0 4.12 0.36 0 0.2 4.54 54 21 3.73 105 1.7 48 81 Bones Fk Cr 77.19 0 8.15 7.63 0 0.39 5.45 47 13 3.41 61 2.2 14 78 Bones Fk Cr 77.19 0 8.15 7.63 0 0.39 4.37 63 21 2.27 82 2.2 19 80 N Pr Anderson Cr 34.55 0 57.99 4.48 0 2.98 5.48 61 22 4.16 77 2.82 25 62 Dial Cr 73.66 0.22 21.54 3.03 0.06 1.49 4.89 52 23 3.85 85 2.9 73 86 Ut Drowning Cr 58.25 0.34 33.46 0.98 0 6.97 4.52 53 20 3.01 88 1.23 24 80 Hubquarter Cr 83.37 0 13.09 2.21 0 1.33 4.96 80 27 4.2 119 2.9 61.8 65 Crooked Fk 70.76 0 25.51 1.35 1.96 0.41 4.76 34 14 3.27 90 2.52 52 79 Little Cr 81.87 0 8.13 0.56 0 9.44 4.78 54 21 4.18 95 2.6 60.5 72 Negro Cr 75.08 0.05 19.48 2.88 0 2.5 4.66 54 20 4.17 74 2.62 87 51 Ut Talbots Cr 88.41 0.04 10.75 0.79 0 0 4.56 52 25 3.94 136 2.4 78 85 MEANS 77.7 0.03 17.3 2.8 0.14 1.8 4.87 56 20.4 3.73 93.2 1.9 52.81 75.82 Reference/Forest Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % BI S EPTS EPTBI EPTN DA Condo Habitat Horsepen Cr 98.26 0 1.65 0.09 0 0 4.18 59 28 3.07 99 0.8 47 83 Moccasin Cr 98.32 0 0.92 0.76 0 0 4.05 56 29 3.54 151 2.5 50 84 Iron Hill Br 80.29 0 17.21 0.97 1.53 0 4.32 70 29 3.78 149 0.9 45.2 87 Ut Hitchcock Cr 47.85 0 19.1 10.76 2.98 19.31 4.35 45 16 2.72 58 1.64 15 67 L Dumas Cr 79.32 0 17.78 1.33 0.07 1.5 4.28 62 30 3.62 163 2 40 90 Millstone Cr 70.25 0.37 15 9.25 0 5.13 4.35 51 18 2.93 93 3 13.9 71 Poison Fk 92.01 0.06 7.49 0.44 0 0 4.11 70 35 3.57 172 2.8 48 84 Joes Cr 78.49 0 14.42 4.47 0 2.62 3.79 44 19 3.13 97 1.8 14 75 Joes Cr 78.49 0 14.42 4.47 0 2.6 4.36 41 19 3.39 90 1.8 12 83 Ut Dobbins Cr 97.16 0 1.56 1.28 0 0 3.59 82 36 2.56 142 0.7 28 76 Wood Run 98.52 0 0.64 0.79 0 0 4 59 25 3.4 170 2.2 44 86 Wood Run 98.52 0 0.64 0.79 0 0.04 3.62 63 36 3.24 179 2.2 46 93 Wood Run 98.52 0 0.64 0.79 0 0 3.94 56 30 3.18 131 2.2 44 86 Ut Drowning Cr 63.24 0.23 25.53 3.3 0 7.7 3.9 75 32 3 125 2.4 21 91 Dutchmans Cr 98.04 0 1.26 0.63 0 0.07 3.3 60 34 2.71 159 2.8 57 86 Dutchmans Cr 98.04 0 1.26 0.63 0 0.1 3.97 74 34 3.51 184 2.8 57 92 Hog Pen Br 99.9 0 0 0 0.1 0 3.75 62 33 2.65 118 0.5 22 85 MEANS 86.6 0.03 8.3 2.4 0.27 2.3 3.99 61 28.4 3.17 134 1.94 35.54 83.47

46

Sites in Blue Indicate “Foothills” data. See Section 13.0.

Table 11. Mountain Study Sites: Raw Data. Raw Data. Five-Tiered Biotic Index Ranges Based on Sorting by Landuse: % Forest, % Developed, % Grass/Pasture, and % Agriculture. Note Overlap of the S, EPTS, EPTN, and EPTBI Between Disturbance Classes. Severe Impact/Developed Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % S EPTSs EPTN EPTBI BI DA Condo Habitat Moore Br 0.33 0 0.66 99.01 0 0 30 2 2 5.5 7.67 0.63 51 48 MEANS 0.33 0 0.66 99.01 0 0 30 2 2 5 7.67 0.63 51 48 Severe Impact/Mostly Developed Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % S EPTs EPTN EPTBI BI DA Condo Habitat Ut French Broad R 17.56 0 15.27 66.66 0.07 0.45 35 7 17 4.34 6.55 0.63 72 78 Ut Mud Cr 27.83 0 2.14 69.82 0 0.21 52 11 40 5.45 6.24 0.9 71 44 Ut Pigeon R 6.55 0 5.35 88.1 0 0 45 9 26 4.39 5.88 0.43 68 52 Ut Pigeon R 30.66 0 59.01 5.92 0 4.4 30 4 8 4.86 6.7 0.37 132 56 MEANS 20.7 0 20.4 57.6 0.02 1.3 41 7.8 23 4.8 6.34 0.58 85.8 57.5 Intermediate Impact/Mixed Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % S EPTs EPTN EPTBI BI DA Condo Habitat Ut French Broad R 37.28 0 22.47 37 0.94 2.31 51 16 55 3.51 5.27 1 81 59 Moore Cr 61.41 0 20.16 18.43 0 0 70 16 77 4.75 5.79 1.1 79 52 Bat Fk 21.73 0.19 30.71 43.54 0.13 3.7 47 9 58 5.78 5.7 1.65 75 45 George Br 46.9 0 34.8 18 0 0 34 9 13 2.42 5.75 0.3 175 69 King Cr 46.18 0 7.82 45.8 0 0.2 36 12 49 4.87 5.56 1.9 45 80 MEANS 42.7 0.04 23.2 32.6 0.21 1.2 48 12 50 4.3 5.61 1.19 91 61 Reference/Mostly Forest Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % S EPTs EPTN EPTBI BI DA Condo Habitat Ross Cr 63.92 0 3.31 32.68 0 0.08 34 15 90 2.94 3.76 1.7 93 84 Mud Cr 78.53 0.33 12.61 7.51 0.04 0.98 52 24 115 3.65 4.25 2.4 48 70 Peter Weaver Cr 74.68 0 12.81 11.72 0 0.79 37 16 78 4.15 4.58 0.9 32 56 Clear Cr 65.05 0 28.46 5.32 0.03 1.14 47 14 60 3.82 4.71 2 41 50 Cox Cr 83.48 0 12.27 3.53 0 0.73 57 16 58 2.84 4.82 2.7 30 73 Mill Cr 54.67 0 37.55 6.34 0.09 1.35 25 11 40 4.54 4.9 1.5 52 37 MEANS 70.1 0.06 17.8 11.2 0.03 0.8 42 16 74 3.7 4.5 1.87 49.3 61.7 Reference/Forest Forest/Wetland % Water % Grass/Pasture % Developed % Barren % Ag % S EPTs EPTN EPTBI BI DA Condo Habitat Ut Wash Cr 97.79 0 2.01 0.2 0 0 67 39 160 1.81 2.46 0.34 14 83 Ledford Br 96.96 0 2.21 0.84 0 0 58 30 177 1.77 2.48 0.45 27 86 Bearwallow Br 94.15 0 0 5.85 0 0 78 42 204 1.44 2.64 0.5 11 85 Buuchanan Cr 87.26 0 2.01 10.73 0 0 69 35 171 1.65 2.39 0.8 61 90 Reedy Br 88.31 0 8.3 3.16 0 0.23 57 36 159 1.67 2.17 0.9 16 88 Reedy Br 88.31 0 8.3 3.16 0 0.23 55 39 184 2.01 2.53 0.9 21 87 UT Stone Mountain Cr 86.2 0 10 3.5 0 0 55 32 155 1.7 2.76 0.5 51 76 Upper Laurel Fk 69.8 0 17.6 12.5 0 0 61 38 181 2.44 2.71 1.1 59 71 George Gap Br 79.7 0 12.2 8 0 0 64 34 142 2.05 2.87 0.7 121 74 Greene Cr 88.8 0 9.6 1.6 0 0 73 41 229 2.42 3.07 1.6 55 65 Harper Cr 84.87 0 7.69 7.05 0.14 0.25 56 26 82 2.68 3.62 1.3 38 80 Ut Double Br 99.51 0 0.49 0 0 0 47 21 102 2.21 2.93 0.3 30 84 Cascade Cr 97.52 0 1.68 0.42 0 0.38 37 18 72 1.18 2.8 1 13 92 Poplar Cr 98.3 0 0.6 1.11 0 0 84 40 185 2.13 2.98 1.2 27 83 Hickory Cr 89.71 0 8.18 2.1 0 0 72 39 163 2.89 3.49 1.2 47 84 Hickory Cr 89.71 0 8.18 2.1 0 0 69 37 173 3.18 3.54 1.2 43 75 Ut Mill Cr 92.96 0 5.1 1.84 0.09 0 72 46 244 2.79 2.99 1.4 28 87 Ut Lower Little R 95.3 0 2.37 2.33 0 0 68 37 176 2.34 2.97 1.5 22 78 Indian Cr 96.4 0.83 1.1 1.61 0 0.06 64 37 156 1.79 2.39 1.7 14 92 Wood Benton Cr 66.91 0 30.25 2.15 0 0.68 74 40 205 2.87 3.51 1.8 57 77 Lambert Fk 98.67 0 1.15 0.18 0 0 65 29 128 2.45 3.42 1.8 29 81 Lynn Br 83.46 0 14.38 2.14 0.02 0 73 39 155 3.25 3.75 2 58 74 Racoon Cr 76.21 0.03 21.83 1.79 0 0.13 73 41 223 3.03 3.66 2.1 63 84 Garrison Cr 96.73 0 1.85 1.41 0 0 72 41 176 2.29 2.72 1.4 18 86 Garrison Cr 96.73 0 1.85 1.41 0 0 55 36 112 1.46 2.12 1.4 16 80 Reedy Br 88.31 0 8.3 0 0 0.23 63 32 167 1.78 2.67 0.9 18 90 Laurel Br 98.86 0 0.17 0.96 0 0 51 28 136 1.47 2.5 1 9 81 Big Bearpen Br 96.97 0 0 3.03 0 0 92 44 266 1.24 2.51 1 12 91 Boyd Br 99.35 0 0 0.65 0 0 53 33 154 2.2 2.64 1 19 79 Bad Fk 95.21 0 0 4.79 0 0 64 40 180 2.08 2.5 2.7 13 92 Lower Cr 98.84 0 0 1.16 0 0 77 46 261 1.94 2.29 1.6 9 97 Singe Cat Br 100 0 0 0 0 0 68 33 166 1.1 2.16 0.8 23 93 Singe Cat Br 100 0 0 0 0 0 49 32 131 1.51 1.9 0.8 24 95 Log Hollow Br 96.45 0 0.35 3.19 0 0 71 44 211 1.22 1.91 0.8 10 94 Stone Mountain Cr 91.4 0 5.3 3 0 0 62 39 216 1.96 2.38 0.9 34 78 Bartlett Cr 84.13 0 2.16 13.7 0 0 61 28 138 0.83 1.94 0.3 46 67 Bartlett Cr 84.13 0 2.16 13.7 0 0 50 23 113 1.05 2.16 0.3 45 84 Roses Cr 98.79 0 0.02 1.19 0 0 64 41 169 1.75 2.71 1.8 22 85 Roses Cr 98.79 0 0.02 1.19 0 0 52 30 151 1.85 2.35 1.8 23 84 Bee Rock Cr 98.51 0 0.16 1.33 0 0 59 36 191 1.33 2.12 1.6 23 89 Bee Rock Cr 98.51 0 0.16 1.33 0 0 66 40 180 2.1 2.6 1.6 21 89 Bee Rock Cr 98.51 0 0.16 1.33 0 0 67 41 202 1.66 2.35 1.6 22 95 Pepper Cr* 89.7 0 4.71 5.58 0 0 58 32 155 1.4 2.23 1 60 77 Cow Cr 98.18 0 0 1.82 0 0 66 37 171 1.39 2.38 1.2 27 98 MEANS 91.9 0.02 4.9 3.1 0 0 64 36 170 1.9 2.67 1.18 31.8 84.1

47

(9.0) Results and Discussion: “Foothills” Versus Piedmont, and Mountain Data One of the only disadvantages in using level III and level IV ecoregion data occurs when the coordinates for a given stream physically places it in a particular level III (e.g., Piedmont) and level IV (e.g., Northern Inner Piedmont) ecoregion yet the vast majority of its watershed actually lies in a different level III (e.g., Mountain) and level IV (e.g., Southern Crystalline Ridges and Mountains) ecoregion. This scenario often causes difficulty in choosing the correct biocriteria thresholds for assessing the stream as defensible arguments can be made for using either Mountain or Piedmont criteria in this instance (NCDWQ 2006). A similar dilemma has been found for a small subset of sites during the course of this study. Unfortunately, this can be a crucial decision since in North Carolina benthic macroinvertebrate metrics have been demonstrated to vary substantially between ecoregions (Lenat 1993). A small portion of the reference streams sampled for this study were arbitrarily deemed as “foothills” sites for the fact that they were located very close to the level III Mountain ecoregion even though the actual sampling point was located in a level III Piedmont ecoregion (Figure 38). Similarly, other sites were physically located in a level III Mountain ecoregion but were surrounded almost entirely by level III Piedmont (Figure 38). Additionally, some of these foothills sites were physically located in the level III Piedmont ecoregion (level IV Northern Inner Piedmont) but portions (to varying degrees) of these streams’ respective watersheds actually included a different level III (Mountain) and level IV ecoregion (Sauratown Mountains; Figure 39). These streams include Cascade Creek and Indian Creek. The remaining foothills sites (Garrison Creek, UT Double Branch, Lambert Fork, Poplar Creek, and Robinette Creek) are both physically located (and whose entire catchment is contained within) the Mountain level III ecoregion (Eastern Blue Ridge Foothills level IV ecoregion) but which are actually encompassed almost entirely by level III Piedmont ecoregion and also have lower elevations relative to other Mountain level III and level IV ecoregion (e.g., Southern Crystalline Ridges and Mountains; Figure 38, Figure 39, Table 12). While these sites are technically located in the level III Mountain ecoregion their extremely close proximity to the level III Piedmont ecoregion make these sites function as ecotonal transitional zones between the Piedmont and Mountains. For example, all of the foothills sites contained core Mountain taxa, and yet some of their BI scores group closer to the Piedmont reference BI data, while others group closer to the Mountain reference data (Figure 40, Table 12). In addition, the foothill reference sites (level IV Ecoregions: Sauratown Mountains, Eastern Foothills, Northern Inner Piedmont; Griffith et al. 2002) all have lower elevations versus sites located within the (level III Mountain) level IV Southern Crystalline Mountains and Ridges, and Southern Metasedimentary Mountains (Figure 40, Table 12; Griffith et al. 2002). This elevation difference suggests why most of the foothills (Sauratown Mountains, Northern Inner Piedmont, Eastern Foothills) reference level IV Piedmont sites have higher BI scores relative to the Southern Crystalline Mountains and Ridges, and Southern Metasedimentary Mountains level IV Mountain reference sites (Figure 40, Table 12), while their close proximity to these level III Mountain ecoregions explain the presence of core Mountain taxa. This elevation gradient also explains why the foothills sites have lower BI scores relative to other (non-foothills) level III and level IV Piedmont data (Figure 42, Table 12). This hypothesis is strongly

48

supported by the data presented in Figure 42, which demonstrates a strong relationship between elevation and biotic index and is consistent with past findings (Eaton and Lenat 1991, Lenat 1993, Cuffney et al. 1997, Cuffney et al. 2000, NCDWQ 2006, Carlisle et al. 2008). Although valid arguments can be made for placing reference foothills sites with either the Piedmont or Mountains reference dataset, the most logical place for these transitional data is with the Mountains reference data. This is recommended primarily due to the fact that if core Mountain taxa are present in a sample, regardless of where that sample was obtained, it is (in nearly all instances) currently rated using Mountain criteria (NCDWQ 2006). The only exceptions to this practice are extreme disjunct occurrences of Drunella walkeri in Coastal Plain reaches of the lower Tar River (Edgecombe and Nash Counties) and Ceratopsyche sparna and Epeorus rubidus from streams located in close proximity to the Uwharrie Mountains in Montgomery and Stanly Counties in the Piedmont. Therefore, since the foothills sites evaluated for this research all contained core Mountain taxa, were all located very close to or actually within level III Mountain ecoregions (and would therefore be rated using Mountain criteria as per NCDWQ 2006), and had BI values most similar to the Mountain values, the foothills data were ultimately deposited within the Mountain dataset.

Figure 38. Level III Ecoregions and “Foothills” Small Stream Study Sites.

49

Figure 39. Level IV Ecoregions and “Foothills” Small Stream Study Sites.

50

Table 12. Level IV and Level III Reference Elevation and Biotic Index Data.

Stream Level III Ecoregions

Level IV Ecoregions County

Elevation (Ft.) BI

“Foothills” Cascade Cr Piedmont Sauratown Mountains Stokes 1018 2.8

UT Mill Cr Piedmont Sauratown Mountains Stokes 891 2.99

Indian Cr Piedmont Sauratown Mountains Stokes 940 2.39

UT Double Br Mountain Eastern Blue Ridge Foothills Burke 1487 2.93

Poplar Cr Mountain Eastern Blue Ridge Foothills Alexander 1266 2.98

Robinette Cr Mountain Eastern Blue Ridge Foothills Alexander 1,250 2.97

Garrison Cr Mountain Eastern Blue Ridge Foothills Burke 1,244 2.72

Garrison Cr Mountain Eastern Blue Ridge Foothills Burke 1,244 2.12

Lambert Fk Mountain Eastern Blue Ridge Foothills Alexander 1,324 3.42

Lynn Br Piedmont Northern Inner Piedmont Stokes 710 3.75

Racoon Cr Piedmont Northern Inner Piedmont Stokes 769 3.66

Hickory Cr Piedmont Northern Inner Piedmont Rockingham 834 3.49

Hickory Cr Piedmont Northern Inner Piedmont Rockingham 834 3.54

Wood Benthon Cr Piedmont Northern Inner Piedmont Stokes 704 3.51

UT Dobbins Cr Piedmont Northern Inner Piedmont Yadkin 1,139 3.59

Mountains

UT Wash Cr Mountain Southern Crystalline Ridges

and Mountains Henderson 2,373 2.46

Ledford Br Mountain Southern Crystalline Ridges

and Mountains Buncombe 2,200 2.48

Bearwallow Br Mountain Southern Crystalline Ridges

and Mountains Transylvania 3,272 2.64

Singe Cat Br Mountain Southern Crystalline Ridges

and Mountains McDowell 2,104 1.9

Singe Cat Br Mountain Southern Crystalline Ridges

and Mountains McDowell 2,104 2.16

Buchanan Cr Mountain Southern Crystalline Ridges

and Mountains McDowell 1,910 2.39

Log Hollow Br Mountain Southern Crystalline Ridges

and Mountains Transylvania 3,068 1.91

Reedy Br Mountain Southern Crystalline Ridges

and Mountains McDowell 1,520 2.17

Reedy Br Mountain Southern Crystalline Ridges

and Mountains McDowell 1,520 2.53

Reedy Br Mountain Southern Crystalline Ridges

and Mountains McDowell 1,520 2.67

Big Bearpen Br Mountain Southern Crystalline Ridges

and Mountains Transylvania 3,047 2.51

Boyd Br Mountain Southern Crystalline Ridges

and Mountains Buncombe 2,191 2.64

Pepper Cr Mountain Southern Crystalline Ridges

and Mountains McDowell 1,851 2.23

Cow Cr Mountain Southern Crystalline Ridges

and Mountains McDowell 1,923 2.38

Bee Rock Cr Mountain Southern Crystalline Ridges

and Mountains McDowell 2,009 2.12

Bee Rock Cr Mountain Southern Crystalline Ridges

and Mountains McDowell 2,009 2.6

Bee Rock Cr Mountain Southern Crystalline Ridges

and Mountains McDowell 2,009 2.35

Lower Cr Mountain Southern Crystalline Ridges

and Mountains Yancey 3,177 2.29

51

Table 12. Level IV and Level III Reference Elevation and Biotic Index Data.

Stream Level III Ecoregions

Level IV Ecoregions County

Elevation (ft) BI

Roses Cr Mountain Southern Crystalline Ridges

and Mountains McDowell 1,825 2.71

Roses Cr Mountain Southern Crystalline Ridges

and Mountains McDowell 1,825 2.35

Bad Fk Mountain Southern Crystalline Ridges

and Mountains Henderson 2,464 2.5

Bartlett Cr Mountain Southern Crystalline Ridges

and Mountains McDowell 2,518 1.94

Stone Mountain Cr Mountain Southern Crystalline Ridges

and Mountains Watauga 2,824 2.16

Laurel Br Mountain Southern Crystalline Ridges

and Mountains McDowell 1,835 2.5Piedmont North Prong Anderson Cr Piedmont Sandhills Harnett 221 4.42North Prong Anderson Cr Piedmont Sandhills Harnett 221 4.68Big Branch Piedmont Sandhills Richmond 317 4.75Millstone Cr Piedmont Sandhills Richmond 379 4.35UT Drowning Cr Piedmont Sandhills Moore 510 4.52W Pr Juniper Br Piedmont Sandhills Scotland 300 4.39UT Hitchcock Cr Piedmont Sandhills Richmond 304 4.35Joes Cr Piedmont Sandhills Richmond 292 3.79Joes Cr Piedmont Sandhills Richmond 292 4.36Bones Fk Cr Piedmont Sandhills Richmond 327 4.37Arnett Branch Piedmont Carolina Slate Belt Montgomery 520 4.5Hogpen Br Piedmont Carolina Slate Belt Montgomery 354 3.75UT Drowning Cr Piedmont Carolina Slate Belt Montgomery 492 3.9Little Cr Piedmont Carolina Slate Belt Montgomery 480 4.78Poisin Fk Piedmont Carolina Slate Belt Montgomery 450 4.11Dutchmans Cr Piedmont Carolina Slate Belt Montgomery 341 3.97Dutchmans Cr Piedmont Carolina Slate Belt Montgomery 341 3.3Wood Run Piedmont Carolina Slate Belt Montgomery 326 4Wood Run Piedmont Carolina Slate Belt Montgomery 326 3.94Wood Run Piedmont Carolina Slate Belt Montgomery 326 3.62Little Dumas Cr Piedmont Carolina Slate Belt Montgomery 567 4.28Crooked Fork Piedmont Carolina Slate Belt Person 531 4.76Dial Creek Piedmont Carolina Slate Belt Durham 440 4.89Iron Hill Branch Piedmont Carolina Slate Belt Montgomery 370 4.32Horsepen Branch Piedmont Carolina Slate Belt Montgomery 328 4.18West Branch Maclean Cr Piedmont Carolina Slate Belt Montgomery 388 4.45West Branch Maclean Cr Piedmont Carolina Slate Belt Montgomery 388 4.54Moccasin Cr Piedmont Carolina Slate Belt Montgomery 342 4.05UT Talbots Cr Piedmont Carolina Slate Belt Randolph 592 4.57UT Bear Swamp Piedmont Northern Outer Piedmont Halifax 153 4.65Hubquarter Creek Piedmont Northern Outer Piedmont Warren 236 4.96Jordan Cr Piedmont Northern Outer Piedmont Warren 210 4.85Tanyard Creek Piedmont Southern Outer Piedmont Davidson 652 4.64Negro Creek Piedmont Southern Outer Piedmont Caswell 577 4.66

52

Figure 40. Biotic Index by Elevation: Level IV Ecoregions of the Mountains and Piedmont.

(10.0) Ecoregions Effects Ecoregions generally denote areas of similarity in ecosystems as well as the type, quality, and quantity of environmental resources (Griffith et al. 2002). Ecoregions are designed to function as a spatial framework for research, assessment, management, and monitoring of ecosystems and components of ecosystems (Griffith et al. 2002). By classifying the spatial differences in the capacities and potential of ecosystems, ecoregions stratify the environment by its probable response to disturbance (Bryce et al. 1999). These general purpose regions are critical for structuring and implementing ecosystem management strategies across federal agencies, state agencies, and nongovernmental organizations that are responsible for different types of resources within the same geographical areas (Omernik et al. 2000). A Roman numeral hierarchical system has been developed for different levels of ecological regions. Level I is the coarsest level dividing North America into 15 ecological regions, level II divides the continent into 52 regions, while level III further divides the continent into 104 ecoregions (Griffith et al. 2002). The level IV ecoregions division is the finest scale and in North Carolina there are 27 level IV ecoregions (Griffith et al. 2002). Previous work in larger North Carolina streams (generally drainage areas greater than 3.0mi2) demonstrated a significant difference in bioclassification thresholds between the Mountain level III, Piedmont level III, and Coastal Plain level III Ecoregions (Lenat 1993). For example, minimum Excellent bioclassification thresholds (based on the BI) in the Mountain level III Ecoregions must

53

be less than 4.00, while minimum Excellent thresholds in the Piedmont level III must be less than 5.14, and in the Coastal Plain level III the minimum is 5.42 (Lenat 1993, NCDWQ, 2006). Relative to the BI scores from larger streams, minimum Excellent BI thresholds in the Mountain level III Ecoregions for small streams were lower (less than or equal to 3.75), and this was also the case for the Piedmont study sites (less than or equal to 4.36). As previous investigations in North Carolina and the southeast have demonstrated, the BI (Lenat 1993) and invertebrate assemblages in general (Feminella 2000) vary substantially by level III Ecoregion. Underlying drivers of this pattern are likely a combination of temperature and elevation (Lenat 1993, Poff 1997, Vinson and Hawkins 1998, Cuffney et al. 2000, Feminella 2000, Hawkins et al. 2000, Carlisle et al. 2008). These findings generally support the physical conditions observed at most of the Mountain and Piedmont reference streams sampled for this study. In general, and particularly in the Mountains, these small headwater streams were 1-2 meters in width with an enclosed (or largely so) riparian canopy. This intense shading, coupled with the naturally close proximity of these headwater streams to their (cooler in temperature) groundwater sources (Alexander et al. 2007) would help explain the lower minimum BI score thresholds established for small Piedmont and Mountain streams relative to thresholds for larger streams in these ecoregions. (11.0) Seasonal Effects Piedmont All of the Piedmont sites were taken from the reference dataset. Wood Run was sampled on three occasions: once in early April, once in early May, and once again in early June. As Figures 53-54 depict, there were no substantial changes in the invertebrate community and there were no changes in the bioclassification between sampling events at this location. West Branch Maclean Creek (Figure 49-50), and Dutchmans Creek (Figure 47-48) were each sampled in early May and then again in early June and there were no changes in bioclassification noted and little variation among the BMI metrics from these months. Hickory Creek (Figure 51-52), Garrison Creek (Figure 45-46), and North Prong Anderson Creek (Figure 43-44) were sampled in early April and then again in early May with no change in bioclassification measured between sampling events and little change among the BMI metrics through time. Jordan Creek was sampled in late April and then again in early June. As was the case with all other Piedmont sites sampled for seasonal effects, there were no changes in bioclassification and little overall change in the BMI metrics temporally (Figure 41-42). Mountain Five Mountain reference sites were examined for seasonal effects: Reedy Branch (Figure 61-62), Singe Cat Branch (Figure 55-56), Bartlett Creek (Figure 57-58), Roses Creek (Figure 59-60), and Bee Rock Creek (Figure 63-64). Singe Cat Branch, Bartlett Creek, and Roses Creek were all sampled in late April and then again two months later in late June and there were no substantial alterations in the invertebrate communities and (with the sole exception of the 6/21/2005 sample at Roses Creek) there were no changes in bioclassification between months at these sites. The 6/21/2005 Roses Creek sample produced a BI of 2.71 and was therefore just 0.01 over the proposed minimum threshold (2.70) for an Excellent bioclassification for Mountain sites. Reedy Branch and Bee Rock

54

Creek were each sampled on three occasions: once in late April, once in late May, and once again in late June. The results of these samples further indicate a stable invertebrate community and none of these samples changed bioclassification. Overall, the Mountain and Piedmont sites exhibited very little change in community metrics between the months examined and (with the exception of the Roses Creek sample which only exceeded the minimum Excellent Mountain bioclassification threshold by 0.01) there were no changes in bioclassification. In summary, there were a total of 11 sites from the Mountains and Piedmont sampled for temporal repeatability between the months of April, May, and June. Only one of these sites (Roses Creek, 6/21/2005) resulted in a different bioclassification from a previous sample. As a result, the temporal repeatability rate of this data set is 91.6%. Consistent temporal repeatability is an important requirement for the validation of biocriteria and previous standards of acceptable levels of temporal repeatability has been set at 90% for estuarine waters (Eaton 2001).

Figures 41-54. Piedmont Seasonality Samples.

Figure 41. Figure 42.

0

1

2

3

4

5

6

BI (4/21/2006) EPTBI (4/21/2006) BI (6/9/2005) EPTBI ( 6/9/2005)

Date

Piedmont Reference Seasonal Comparison: BI and EPTB I (Jordan Creek)

0

10

20

30

40

50

60

70

80

90

S (4/21/2006) EPTS (4/21/2006) EPTN (4/21/2006) S (6/9/2005) EPTS (6/9/2005) EPTN (6/9/2005)

Date

Piedmont Reference Seasonal Comparison: S, EPTS, an d EPTN (Jordan Creek)

55

Figures 41-54. Piedmont Seasonality Samples (Continued).

Figure 43.

Figure 44.

Figure 45.

Figure 46.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

BI (4/5/2006) EPTBI (4/5/2006) BI (5/16/2005) EPTBI (5 /16/2005)

Date

Piedmont Reference Seasonal Comparison: BI and EPTB I (North Pr. Anderson Cr)

0

10

20

30

40

50

60

70

80

S (4/5/2006) EPTS (4/5/2006) EPTN (4/5/2006) S (5/16/2005) EPTS (5/16/2005) EPTN (5/16/2005)

Date

Piedmont Reference Seasonal Comparison: S, EPTS, and EPTN (N Pr Anderson Cr)

0

0.5

1

1.5

2

2.5

3

BI (4/19/2005) EPTBI (4/19/2005) BI (5/22/2006) EPTBI (5/22/2006)

Date

Piedmont Reference Seasonal Comparison: BI and EPTB I (Garrison Creek)

0

20

40

60

80

100

120

140

160

180

S (4/19/2005) EPTS(4/19/2005)

EPTN(4/19/2005)

S (5/22/2006) EPTS(5/22/2006)

EPTN(5/22/2006)

Date

Piedmont Reference Seasonal Comparison: S, EPTS, an d EPTN (Garrison Creek)

56

Figures 41-54. Piedmont Seasonality Samples (Continued). Figure 47. Figure 48. Figure 49. Figure 50.

0

0.5

1

1.5

2

2.5

3

3.5

4

BI (5/11/2005) EPTBI (5/11/2005) BI (6/16/2005) EPTBI (6/16/2005)

Date

Piedmont Reference Seasonal Comparison: BI and EPTB I (Dutchmans Creek)

0

20

40

60

80

100

120

140

160

180

200

S (5/11/2005) EPTS (5/11/2005) EPTN (5/11/2005) S (6/16/2005) EPTS(6/16/2005) EPTN(6/16/2005)

Date

Piedmont Reference Seasonal Comparison: S, EPTS, an d EPTN (Dutchmans Creek)

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

BI (5/3/2005) EPTBI (5/3/2005) BI (6/6/2005) EPTBI (6/ 6/2005)

Date

Piedmont Reference Seasonal Comparison: BI and EPTB I (W Br Maclean Cr)

0

20

40

60

80

100

120

S (5/3/2005) EPTS (5/3/2005) EPTN (5/3/2005) S (6/6/2005) EPTS(6/6/2005) EPTN(6/6/2005)

Date

Piedmont Reference Seasonal Comparison: S, EPTS, an d EPTN (W Br Maclean Cr)

57

Figures 41-54. Piedmont Seasonality Samples (Continued). Figure 51. Figure 52. Figure 53. Figure 54.

0

0.5

1

1.5

2

2.5

3

3.5

4

BI (4/18/2006) EPTBI (4/18/2006) BI (5/31/2005) EPTBI (5/31/2005)

Date

Piedmont Reference Seasonal Comparison: BI and EPTB I (Hickory Creek)

0

20

40

60

80

100

120

140

160

180

S (4/18/2006) EPTS(4/18/2006)

EPTN(4/18/2006)

S (5/31/2005) EPTS(5/31/2005)

EPTN(5/31/2005)

Date

Piedmont Reference Seasonal Comparison: S, EPTS, an d EPTN (Hickory Creek)

0

0.5

1

1.5

2

2.5

3

3.5

4

BI (4/7/2006) EPTBI(4/7/2006)

BI (5/11/2005) EPTBI(5/11/2005)

BI (6/16/2005) EPTBI(6/16/2005)

Date

Piedmont Reference Seasonal Comparison: BI and EPTB I (Wood Run)

0

20

40

60

80

100

120

140

160

180

S(4/7/2006)

EPTS(4/7/2006)

EPTN(4/7/2006)

S(5/11/2005)

EPTS(5/11/2005)

EPTN(5/11/2005)

S(6/16/2005)

EPTS(6/16/2005)

EPTN(6/16/2005)

Date

Piedmont Reference Seasonal Comparison: S, EPTS, an d EPTN (Wood Run)

58

Figures 55-64. Mountain Seasonality Samples. Figure 55. Figure 56. Figure 57. Figure 58. .

0

0.5

1

1.5

2

2.5

BI (4/21/2005) EPTBI (4/21/2005) BI (6/22/2005) EPTBI (6/22/2005)

Date

Mountain Reference Seasonal Comparison: BI and EPTB I (Singe Cat Creek)

0

20

40

60

80

100

120

140

160

180

S (4/21/2005) EPTS(4/21/2005)

EPTN(4/21/2005)

S (6/22/2005) EPTS(6/22/2005)

EPTN(6/22/2005)

Date

Mountain Reference Seasonal Comparison: S, EPTS, an d EPTN (Singe Cat Creek)

0

0.5

1

1.5

2

2.5

BI (4/20/2005) EPTBI (4/20/2005) BI (6/22/2005) EPTBI (6/22/2005)

Date

Mountain Reference Seasonal Comparison: BI and EPTB I (Bartlett Creek)

0

20

40

60

80

100

120

140

S (4/20/2005) EPTS(4/20/2005)

EPTN(4/20/2005)

S (6/22/2005) EPTS(6/22/2005)

EPTN(6/22/2005)

Date

Mountain Reference Seasonal Comparison: S, EPTS, an d EPTN (Bartlett Creek)

59

Figures 55-64. Mountain Seasonality Samples (Continued).

Figure 59. Figure 60. Figure 61. Figure 62.

0

0.5

1

1.5

2

2.5

3

BI (4/20/2005) EPTBI (4/20/2005) BI (6/21/2005) EPTBI (6/21/2005)

Date

Mountain Reference Seasonal Comparison: BI and EPTB I (Roses Creek)

0

20

40

60

80

100

120

140

160

180

S (4/20/2005) EPTS(4/20/2005)

EPTN(4/20/2005)

S (6/21/2005) EPTS(6/21/2005)

EPTN(6/21/2005)

Date

Mountain Reference Seasonal Comparison: S, EPTS, an d EPTN (Roses Creek)

0

0.5

1

1.5

2

2.5

3

BI (4/21/2005) EPTBI(4/21/2005)

BI (5/22/2005) EPTBI(5/22/2005)

BI (6/22/2005) EPTBI(6/22/2005)

Date

Mountain Reference Seasonal Comparison: BI and EPTB I (Reedy Branch)

0

20

40

60

80

100

120

140

160

180

S(4/21/2005)

EPTS(4/21/2005)

EPTN(4/21/2005)

S(5/22/2005)

EPTS(5/22/2005)

EPTN(5/22/2005)

S(6/22/2005)

EPTS(6/22/2005)

EPTN(6/22/2005)

Date

Mountain Reference Seasonal Comparison: S, EPTS, an d EPTN (Reedy Branch)

60

Figures 55-64. Mountain Seasonality Samples (Continued). Figure 63. Figure 64. (12.0) Conclusions Regardless of ecoregion or disturbance class, the BI had the lowest CV relative to the EPTS, EPTBI, EPTN, and S, and was also the most strongly correlated to two of the most important determinants on water quality and the biological integrity of benthic macroinvertebrate communities: percent forest/wetland and percent developed. In addition, the BI was essentially equivalent in performance to the EPTBI among Mountain sites in response to percent agriculture, percent sand, percent sand+silt, and habitat score, but significantly outperformed the EPTBI in the Mountains when it came to correlation to percent grass/pasture, ag+grass/pasture, and conductivity and in the Piedmont the BI was most strongly correlated to habitat and was comparable with the EPTBI in terms of correlation to conductivity. Moreover, the BI exhibited no overlap among the coarse or fine-scaled disturbance classes in the Mountains or Piedmont. Conversely, in both the Piedmont and Mountains, the S, EPTN, EPTS, and EPTBI demonstrated significant overlap between the landuse derived coarse-scale, and fine-scale disturbance classes. Consequently, of the five BMI metrics examined for this study only the BI was capable of reliably distinguishing between the worst and best sites, as well as between sites with only subtle differences in disturbance gradients.

0

0.5

1

1.5

2

2.5

3

BI (4/20/2005) EPTBI(4/20/2005)

BI (5/22/2005) EPTBI(5/22/2005)

BI (6/21/2005) EPTBI(6/21/2005)

Date

Mountain Reference Seasonal Comparison: BI and EPTB I (Bee Rock Creek)

0

50

100

150

200

250

S(4/20/2005)

EPTS(4/20/2005)

EPTN(4/20/2005)

S(5/22/2005)

EPTS(5/22/2005)

EPTN(5/22/2005)

S(6/21/2005)

EPTS(6/21/2005)

EPTN(6/21/2005)

Date

Mountain Reference Seasonal Comparison: S, EPTS, an d EPTN (Bee Rock Creek)

61

(13.0) Future Work Although this is a large dataset, it is not comprehensive. To this end, there are some disturbance classes within the level IV and level III ecoregions that are (to varying degrees of severity) somewhat underrepresented here. These would include (in approximate order of priority): 1) severe impact streams in the Mountains , 2) intermediate impact sites in the Mountains, 3)severe impact streams in the Sandhills level IV Ecoregions, and 4) reference Piedmont samples, particularly outside of the Uwharrie National Forest area. In terms of additional seasonal data, repeat sampling from a subset of reference Piedmont and Mountain sites in March and (to a lesser extent) June as well as July and August in an effort to expand the small streams sampling window. In addition, it should be the ultimate goal of this methodology to evaluate how these invertebrate communities vary throughout the calendar year so that these systems could eventually be sampled at anytime. While there was a good overall range of drainage areas sampled in this study, additional effort should be focused on streams that are < 0.3mi2. However, accurately finding such sites can be difficult due to poor map resolution and limitations to GIS-accuracy at these very small catchment scales. Acknowledgements This work could not have been completed without the collection and sample processing support of the following BAU members: Steve Beaty, Deirdre Black, Bill Crouch, Trish MacPherson, Blair Prusha, Michelle Simonson, Cathy Tyndall, and Mike Walters. Additional gratitude is directed to Dave Lenat, Larry Eaton, Neil Medlin, Ken Eagleson, and Dave Penrose for their collective efforts towards the establishment of earlier biocriteria for other North Carolina lotic systems. Additional thanks go out to Mark Hale for GIS support, Steve Kroeger for helpful edits to the graphical format of the statistical figures and Dr. Catherine Truxillo who provided detailed statistical guidance. Final acknowledgement must be bestowed upon Jimmie Overton and Trish MacPherson for allocating the necessary resources to complete this project. References Alexander, R B., Boyer, E.W., R.A., Smith, G.E., Schwarz, and R.B., Moore. 2007. The

role of headwater streams in downstream water quality. Journal of the American Water Resources Association. 43(1): 41-59.

American Rivers. 2008. Stream Miles by State: www.americanrivers.org/site /DocServer/Stream_Miles_Table_FINAL__2_.pdf?docID=4081 Angradi, T.R. 1999. Fine sediment and macroinvertebrate assemblages in Appalachian

streams: a field experiment with biomonitoring applications. Journal of the North American Benthological Society 18: 49-66.

Black, R.W., and M.D. Munn. 2004. Using macroinvertebrates to identify biota-land

cover optima at multiple scales in the Pacific Northwest, USA. Journal of the North American Benthological Society 23(2): 340-362.

62

Bressler, D.W., M.J. Paul, A.H. Purcell, M.T. Barbour, E. T. Rankin, and V.H. Resh. 2009. Assessment tools for urban catchments: developing stressor gradients. Journal of the American Water Resources Association 45(2):291-305.

Brinson, M.M. 1993. Changes in the functioning of wetlands along environmental

gradients. Wetlands 13, 65-74. Bryce, S.A., Omernik, J.M. and D.P. Larsen. 1999. Ecoregions—a geographical

framework to guide risk characterization and ecosystem management: Environmental Practice, v. 1, no. 3, p. 141-155.

Bryce, S.A., D.P. Larsen, R.M. Hughes, and P.R. Kaufman. 1999 Assessing relative risks

to aquatic ecosystems: a mid-Appalachian case study. Journal of the American Water Resources Association 35(1): 23-36.

Burton, J.G. 2004. Headwater catchments: Estimating surface drainage extent across

North Carolina and correlations between landuse, near stream, and water quality indicators in the Piedmont physiographic region. North Carolina State University PhD Dissertation. Raleigh, North Carolina, 92 pp.

Cao, Y., C.P. Hawkins, and M.R. Vinson. 2003. Measuring and controlling data quality

in biological assemblage surveys with special reference to stream benthic macroinvertebrates. Freshwater Biology 48:1898-1911.

Carlisle, D.M., C.P. Hawkins, M.R. Meador, M. Potapova and J. Falcone. 2008.

Biological assessments of Appalachian streams based on predictive models for fish, macroinvertebrates, and diatom assemblages. Journal of the North American Benthological Society 27(1): 16-37.

Carter, J.L. and V.H. Resh. 2001. After site selection and before data analysis: sampling,

sorting, and laboratory procedures used in stream benthic macroinvertebrate monitoring programs by USA state agencies. Journal of the American Benthological Society 20: 658-676.

Chutter, F.M. 1972. An empirical biotic index of the quality of water in South African

streams and rivers. Water Research 6:19-30. Clarke, R.T. and D. Hering. 2006. Errors and uncertainty in bioassessment methods—

major results and conclusions from the STAR project and their application using STARUGS. Hydrobiologia 566:433-439.

Clarke, R.T., A. Lorenz, L. Sandin, A. Schmidt-Kloiber, J. Strackbein, N.T. Kneebone,

and P. Haase. 2006. Effects of sampling and sub-sampling variation using the STAR-AQEM sampling protocol on the precision of macroinvertebrate metrics. Hydrobiologia 566:441-459.

63

Crawford, J.K., and D.R. Lenat. 1989. Effects of landuse on the water quality and biota of three streams in the Piedmont province of North Carolina. Water Resources Investigations Report 89-4007, US Geological Survey, Raleigh, NC, USA.

Connolly, N.M., and R.G. Pearson. 2007. The effect of fine sediment on tropical stream

macroinvertebrate assemblages: a comparison using flow-through artificial stream channels and recirculating mesocosms. Hydrobiologia 592: 423-438.

Cover, M.R., C.L. May, W.E. Dietrich, and V.H. Resh. 2007. Quantitative linkages

among sediment supply, streambed fine sediment, and benthic macroinvertebrates in northern California streams. Journal of the North American Benthological Society 27(1):135-149.

Crunkilton, R.L. and R.M. Duchrow. 1991. Use of stream order and biological indices to

assess water quality in the Osage and Black river basins of Missouri. Hydrobiologia 223: 155-166.

Cuffney, T.F., M.R., Meador, S.D. Porter, and M.E. Gurtz. 1997. Distribution of fish,

benthic invertebrate, and algal communities in relation to physical and chemical conditions, Yakima River Basin, Washington, 1990, United States Geological Survey Water Resources Investigations Report 96-4280.

Cuffney, T.F., M.R. Meador, S.D. Porter, and M.E. Gurtz. 2000. Responses of physical,

chemical and biological indicators of water quality to a gradient of agricultural land use in the Yakima river basin, Washington. Environmental Monitoring and Assessment. 64:259-270.

Cuffney, T.F., M.D. Bilger, and A.M. Haigler. 2007. Ambiguous taxa: effects on the

characterization and interpretation of invertebrate assemblages. Journal of the North American Benthological Society 26(2):286-307.

Diamond, J.M., M.T. Barbour, and J.B. Stribling. 1996. Characterizing and comparing

bioassessment approaches and their results: a perspective. Journal of the North American Benthological Society 15: 713-727.

Eaton, L.E., and D.R. Lenat. 1991. Comparison of a rapid bioassessment method with

North Carolina’s macroinvertebrate collection method. Journal of the North American Benthological Society 10:335-338.

Eaton, L.E. 2001. Development and validation of biocriteria using benthic

macroinvertebrates for North Carolina estuarine waters. Marine Pollution Bulletin. Vol. 42, No. 1 pp. 23-30.

Feminella, J.W., 2000. Correspondence between stream macroinvertebrate assemblages

and four ecoregions of the southeastern USA. Journal of the North American Benthological Society. 19(3):442-461.

64

Flotemersch, J.E., J.B. Stribling, and M.J. Paul. 2006. Concepts and approaches for the

bioassessment of non-wadeable streams and rivers. EPA/600/R-06/127. Office of Research and Development, US Environmental Protection Agency, Cincinnati, Ohio.

Gelber, J.B. 2004. Mesoscale spatial variability of selected aquatic invertebrate community metrics from a minimally impaired stream segment. Journal of the North American Benthological Society 23(3):616-633.

Gomez, K.A., and A.A. Gomez. 1976. Statistical procedures for agricultural research, 2nd

ed. John Wiley and Sons, New York. Griffith, G.E., Omernik, J.M., Comstock, J.A. Schafale, M.P., McNab, W.H., Lenat,

D.R., MacPherson, T.F., Glover, J.B. and Shelburne, V.B. 2002. Ecoregions of North Carolina and South Carolina. (Color poster with map, descriptive text, summary tables and photographs): Reston, VA, U.S. Geological Survey (map scale 1:1,500,000).

Hawkins, C.P., R.H. Norris, J.N. Hogue, and J.W. Feminella. 2000. Development and

evaluation of predictive models for measuring the biological integrity of streams. Ecological Applications 10:1456-1477.

Herbst, D.B., and E L. Silldorff. 2006. Comparison of the performance of different

bioassessment methods: similar evaluations of biotic integrity from separate programs and procedures. Journal of the North American Benthological Society 25(2): 513-530.

Herlihy, A.T., J.L. Stoddard and C.B. Johnson. 1998. The relationship between stream

chemistry and watershed land use data in the Mid-Atlantic Region, U.S. Water, Air and Soil Pollution 105:377-386.

Hilsenhoff, W.L. 1977. Use of arthropods to evaluate water quality in streams. Wisconsin

Department of Natural Resources Technical Bulletin No. 100. Hilsenhoff, W.L. 1982. Using a biotic index to evaluate water quality in streams.

Wisconsin Department of Natural Resources Technical Bulletin No. 132. Hilsenhoff, W.L. 1987. An improved biotic index of organic stream pollution. Great

Lakes Entomologist 20:31-39. Hynes, H.B.N. 1970. The Ecology of Running Waters. Univ. Toronto Press, 555 pp. Jackson, S. 2004. Using biological assessments to refine designated aquatic life uses:

EPA/State workgroup. National Biological Assessment and Criteria Workshop, Coeur d’Alene, Idaho. (Available from: http://www.epa.gov/waterscience/biocriteria/modules/).

65

Karr, J.R., K.D. Fausch, P.L. Angermeir, P.R. Yant and I.J. Schlosser. 1986. Assessing

biological integrity in running waters: a method and its rationale. Illinois Natural History Survey (Special Publication) 5, 28 pp.

Karr, J.R. 1991. Biological integrity: a long-neglected aspect of water resource

management. Ecological Applications 1: 66-84. Karr, J.R. 1993. Defining and assessing ecological integrity: beyond water quality.

Environmental Toxicology and Chemistry 12: 1521-1531. Karr, J.R. and C.O. Yoder. 2004. Biological assessment and criteria improve TMDL

planning and decision-making. Journal of Environmental Engineering 130:594-604.

Kerans, B.L., J.R. Karr, and S.A. Ahlstedt. 1992. Aquatic invertebrate assemblages:

spatial and temporal differences among sampling protocols. Journal of the North American Benthological Society 11:377-390.

Kondratieff, B., R. Voshell. 1980. Life history and ecology of Stenonema modestum

(Banks) (Ephemeroptera: Heptageniidae) in Virginia, USA. Aquatic Insects 2, 177-189.

Kratzer, E.B., J.K. Jackson, D.B. Arscott, A.K. Aufdenkampe, C.L. Dow, L.A. Kaplan,

J.D. Newbold, and B.W. Sweeney. 2006. Journal of the North American Benthological Society 25(4): 954-976.

Larsen, M., D. Booth, and S. Morley. 2001. Effectiveness of large woody debris in

stream rehabilitation projects in urban basins. Ecological Engineering. 18, 211-226.

Lenat, D.R. 1988. Water quality assessment of streams using a qualitative collection

method for benthic macroinvertebrates. Journal of the North American Benthological Society 7, 222-233.

Lenat, D. R. 1993. A biotic index for the southeastern United States: derivation and list

of tolerance values, with criteria for assigning water-quality ratings. Journal of the North American Benthological Society 12(3): 279-290.

66

Lenat, D.R., and M.T. Barbour. 1994. Using benthic macroinvertebrate community structure for rapid, cost-effective eater quality monitoring: rapid bioassessment. Pages 187-215 In S.L. Loeb and A. Spacie, editors. Biological Monitoring of Aquatic Systems. Lewis, Boca Raton, Florida, USA.

Leopold, L.B. 1956. Ephemeral Streams: Hydraulic Factors and Their Relation to

Drainage Net. U.S. Geological Survey Professional Paper, 282-A, 38p. Leopold, L.B., M.G. Wolman, and T.P. Miller. 1964. Fluvial Processes in

Geomorphology. W.H. Freeman and Company, San Francisco. Lowrance, R., L. Altier, J. Newbold, R. Schnabel, P. Groffman, J. Denver, D. Correll, J.

Gilliam, J. Robinson, J. Brinsfield, R. Staver, K. Lucas, W. Todd. 1997. Water quality functions of riparian forest buffers in Chesapeake Bay watersheds. Environmental Management. 21, 687-712.

Maloney, K.O., J.W. Feminella, and J.P. Mulholland. 2002. Effects of watershed

disturbance on macroinvertebrate communities in small streams at Fort Benning, GA. Journal of the North American Benthological Society21(3): 487-501.

Meador, M.R. and R.M. Goldstein. 2003. Assessing water quality at large geographic

scales: relations among land use, water physiochemistry, riparian condition, and fish community structure. Environmental Management. 31: 504-517.

Meyer, J.L., D.L. Strayer, J.B. Wallace, S.L. Eggert, G.S. Helfman, and N.E. Leonard.

2007. The contribution of headwater streams to biodiversity in river networks. Journal of the American Water Resources Association. 43(1): 86-103.

Moore, A.A., and M.A. Palmer. 2005. Macroinvertebrate biodiversity in agricultural and

urban headwater streams: implications for conservation and management. Ecological Applications 15:1169-1177.

Narf, R.P., E.L. Lange, and R.C. Wildman. 1984. Statistical procedures for applying

Hilsenhoff’s Biotic Index. Journal of Freshwater Ecology 2:441-448. NCDWQ, 2006. Standard Operating Procedures for Benthic Macroinvertebrates. NC

Dept. of Environment and Natural Resources, Division of Water Quality, Water Quality Section, Environmental Sciences Branch. July, 2006.

Odum, E.P., J.T. Finn, and E.H. Franz. 1979. Perturbation theory and the subsidy-stress

gradient. Bioscience 29: 349-352. Omernik, J.M., S.S. Chapman, R.A. Lillie, and R.T., Dumke. 2000. Ecoregions of

Wisconsin: Transactions of the Wisconsin Academy of Sciences, Arts and Letters, v. 88, no. 2000, p. 77-103.

67

Paul, M.J., and J.L. Meyer. 2001. Streams in the urban landscape. Annual Review of Ecology, Evolution, and Systematics 32:333-365.

Plafkin, J.L., M.T. Barbour, S.D. Porter, S.K. Gross, and R.M. Hughes. 1989. Rapid

bioassessment protocols for use in streams and rivers: benthic macroinvertebrates and fish. US Environmental Protection Agency Office of Water, EPA/4441/4-89-001.

Poff, N.L. 1997. Landscape filters and species traits: towards mechanistic understanding

and prediction stream ecology. Journal of the North American Benthological Society 16:391-409.

Quinn, J.M. 2000. Effects of pastoral development. In New Zealand Stream

Invertebrates: Ecology and Implications for Management, ed. KJ Collier, MJ Winterbourn, pp. 208-229. Christchurch, NZ: Caxton.

Reid, J. 2008. State Geologist. North Carolina Geological Survey, Raleigh, NC 27615,

USA. Personal Communication. Rehn, A. C., P.R. Ode, C.P. Hawkins. 2007. Comparison of targeted-riffle and reach-

wide benthic macroinvertebrate samples: implications for data sharing in stream-condition assessments. Journal of the North American Benthological Society 26(2): 332-348.

Rheinhardt, R.R., M.C. Rheinhardt, M.C. Brinson, K.E. Faser. 1999. Application of

reference data for assessing and restoring headwater ecosystems. Ecological Restoration 7, 241-251.

Rheinhardt, R., M. Brinson, R. Christian, K. Miller, G. Meyer, C. Bason, E. Hardison.

2005. Development of ecological assessments for planning Coastal Plain stream restoration in coastal North Carolina. Report to the North Carolina Ecosystem Enhancement Program. North Carolina Department of Enviroment and Natural Resources, Raleigh, NC USA.

SAS Institute Inc., Introduction to SAS for Biostatisticians and Epidemiologists. 2005.

SAS Institute Inc., Cary, NC 27513, USA. SAS Institute Inc., JMP Statistical Software. Version 8.0. SAS Institute Inc., Cary, NC

27513, USA. Schloesser, D.W., T.F. Nalepa. 2002. Comparison of five benthic samplers to collect

burrowing mayfly nymphs (Hexagenia Spp.: Ephemeroptera: Ephemerida) in sediments of the Laurentian Great Lakes. Journal of the North American Benthological Society 21(3): 487-501.

68

Shields Jr., F.D., S.S. Knight, C.M. Cooper. 1995. Use of the index of biotic integrity to assess physical habitat degradation in warmwater streams. Hydrobiologia 312, 191-208. 1995.

Smart, M.M., J.R. Jones, and J.L. Sebaugh. 1985. Stream-watershed relations in the

Missouri Ozark Plateau Province. Journal of Environmental Quality 14(1): 77-82 Smith, R.F., and W. O. Lamp. 2007. Comparison of insect communities between adjacent

headwater and main-stem streams in urban and rural watershed. Journal of the North American Benthological Society 27(1): 161-175.

Stewart, A.J., and J.M. Loar. 1994. Spatial and temporal variation in biomonitoring data.

Pages 91-124 In S.L. Loeb and A. Spacie, editors. Biological Monitoring of Aquatic Systems. Lewis, Boca Raton, Florida, USA.

Strayer, D.L., R.E. Beighley, L.C. Thompson, S. Brooks, and C. Nilsson. 2003. Effects of

land cover on stream ecosystems: roles of empirical models and scaling issues. Ecosystems 6: 407-423.

Stribling, J.B., B.K., Jessup, and D.L. Feldman. 2008. Precision of benthic

macroinvertebrate indicators of stream condition in Montana. Journal of the North American Benthological Society 27(1): 58-67.

Swift, L.W., Jr. 1983. Duration of stream temperature increases following forest cutting

in the southern Appalachian Mountains. Pages 273-275 In Proceedings of the International Symposium of Hydrometerology, Denver, Colorado, USA, 13-17 June 1982. American Water Resources Association, Bethesda, Maryland, USA.

Thomson, J.R., D.D. Hart, D.F. Charles, T.L. Nightengale, and D.M. Winter. 2004.

Effects of removal of a small dam on downstream macroinvertebrates and algal assemblages in a Pennsylvania stream. Journal of the North American Benthological Society 24(1):192-207.

Truxillo, C., Ph.D. 2008. Biostatistician, SAS Institute Inc., Cary, NC 27513, USA. Personal Communication. Tullos, D.D., D.L. Penrose, and G.D. Jennings. 2006. Development and application of a

bioindicator for benthic habitat enhancement in the North Carolina Piedmont. Ecological Engineering 27:228-241.

Vinson, M.R., and C.P. Hawkins. 1998. Biodiversity of stream insects: variation at local,

basin, and regional scales. Annual Review of Entomology 43:271-293. Wallace, J.B., J.W. Grubaugh, and M.R. Whiles. 1996. Biotic indices and stream

ecosystem processes: results from an experimental study. Ecological Applications 6(1):140-151.

69

Watzin, M.C. and A.W. McIntosh. 1999. Aquatic ecosystems in agricultural landscapes:

a review of ecological indicators and achievable ecological outcomes. J. Soil Water Conserv. 54: 636-644.

Weisburg, S.B. J.A. Ranasinghe, D. M. Dauer, L.C. Schaffner, R.J. Diaz, and J.B.

Frithsen. 1997. An estuarine benthic index of biotic integrity (B-IBI) for Chesapeake Bay. Estuaries 20, 149-158.

Weigel, B.M. 2002. Development of stream macroinvertebrate models that predict

watershed and local stressors in Wisconsin. Journal of the North American Benthological Society. 22(1): 12-142.

Whittier, T.R., J.L. Stoddard, D.P. Larsen, and A.T. Herlihy. 2007. Selecting reference

sites for stream biological assessments: best professional judgment or objective criteria. Journal of the North American Benthological Society. 26(2): 349-360.

Wipfli, M.S., J.S. Richardson, and R.J. Naiman. 2007. Ecological linkages between

headwaters and downstream ecosystems: transport of organic matter, invertebrates, and wood down headwater channels. Journal of the American Water Resources Association. 43(1): 72-85.

Yoder, C.O., and E.T. Rankin. 1998. The role of biological indicators in a state water

quality management process. Environmental Monitoring and Assessment 51:61-88.

Zweig, L.D., and C.F. Rabeni. 2001. Biomonitoring for deposited sediment using benthic

invertebrates: a test on 4 Missouri streams. . Journal of the North American Benthological Society 20(4):643-657.